Beyond Traditional Device Intelligence: The Future of Digital Identity & Fraud Prevention

Register Now →

Solutions

Industries

Customers

Partners

Resources

Request a Demo →

Webinar: Beyond Traditional Device Intelligence

Register Now →

Webinar: Beyond Traditional Device Intelligence

Register Now →

Oscilar Founder Neha Narkhede: Eliminating Risk With AI

Oscilar Founder Neha Narkhede: Eliminating Risk With AI

Amy Sariego

Read time:

20 minutes

•

Feb 12, 2025

Content

Share this article
Subscribe to updates

This week, our CEO & Co-founder, Neha Narkhede, joined Nik Milanović on This Week in Fintech for an in-depth discussion about the future of financial risk management and Oscilar's innovative approach to AI-powered risk decisioning.

Key Highlights

In this conversation, Neha explains how a unified platform approach to AI Risk Decisioning creates a comprehensive view of fraud, credit, and compliance risks. She discusses:

  • The critical gaps in today's fragmented risk management landscape and why point solutions are insufficient for modern financial institutions

  • How Oscilar's unified platform enables real-time risk decisions by combining fraud detection, compliance monitoring, and credit decisioning

  • The transformative impact of shared risk signals across the customer lifecycle

  • Real customer success stories, including how Trans Pecos Banks achieved 40% reduction in AML operations costs and 70% reduction in alert review time

  • The growing importance of AI in combating sophisticated fraud techniques, from synthetic identities to deep fakes

  • The increasing regulatory scrutiny of sponsor banks and why real-time compliance monitoring is crucial

  • Her vision for the future of risk management, including hyper-personalized decisioning and unified compliance platforms

For risk operations professionals, compliance teams, and fintech leaders, this conversation provides valuable insights into building more resilient financial systems through unified, AI-powered risk management.

You can listen to the full episode here

Full Episode Transcript

Nik Milanovic: Thank you everyone for joining us today. We are back for another This Week in Fintech podcast, and I am very excited for our guest today. Our guest is a founder, a repeat founder, I should say, who has been in the financial services and fintech space for a while, and I'm very excited to hear from her. Her name is Neha Narkhede, and she is the founder of Oscilar.

For those of you who may not yet be familiar with Oscilar, Oscilar is an AI risk decisioning platform that helps organizations manage onboarding, fraud, credit and compliance risks.

Neha is a repeat founder who spent a while within the space building out new products, and we're going to dig in today to what Oscilar is building, what their vision is for the future, a few of their customer case studies and what comes next.

Thanks again for joining us and Neha, thank you so much for making the time to be with us.

Neha Narkhede: Thank you, Nik. Very excited to be here, just kicking off. I know we've done a few events together This Week in Fintech and Oscilar. We've partnered in the UK. We've partnered in the US. We've partnered in Brazil. It feels like you're everywhere as Oscilar is growing around the world. Oscilar is a global company, and we're a fast growing business that operates in North America, Europe, Latin and also in the Middle East.

Nik Milanovic: For those who don't know, Neha previously founded a company called Confluent, where I believe you were there for ten years as their CTO chief product officer, and then later on a board member, building a company for real time data and Apache Kafka. I'm curious, could you walk us through your journey from founding Confluent in Kafka to founding Oscilar?

Neha Narkhede: It's been a really fun journey at Confluent. I co-founded Confluent to build a real time data streaming platform powered by a very popular open source system called Apache Kafka, which I was one of the original creators for, and we help businesses all around the world harness real time data and AI to transform their operations.

While working with the leading banks, fintechs, merchants, travel and hospitality, I saw firsthand the fraud and risk management were among the most critical and high impact use cases of real time data and AI in the modern world. However, the existing tools for agile, real time, fraud prevention, risk and compliance, they fell short in a number of ways, and people were trying to work around it and trying to build various tools around it, which actually didn't make a lot of sense.

People are seeing more and more transactions of value moving online at a very fast pace since the COVID pandemic, and they still are without modern and sophisticated risk infrastructure. So recognizing this massive trend and really driven by my passion to secure online transactions, I co-founded Oscilar to empower businesses to detect fraud, manage compliance, make better credit risk decisioning with cutting edge data analysis and machine learning all through one unified risk platform, which is being done for the very first time.

Nik Milanovic: And I noticed that when you founded and announced Oscilar, something that was interesting to me was that you and your co-founder, Sasha Kulkarni also self-funded it to get it off the ground, which to me, stands out as clear example of high conviction in the product you're building, that this is something that you're willing to self fund just to get up and off the ground.

You mentioned while you were at Confluent, you were working with a set of risk tools - tools that you just didn't feel really met your needs, and so it kind of started this burning desire to actually build a better risk product based on your own experience. I'm curious, what were the moments at Confluent that made you realize that there was a real gap in the market and opportunity for building a tool like this? What moments gave you such high conviction that you were willing to start over from the ground up as a repeat entrepreneur and self fund to get this product off the ground and up and running?

Neha Narkhede: Given how big the company is, I saw companies across different industry verticals struggle with the same problems. And the trend I saw is, every industry vertical nearly has some online transactions of some type, and any online transaction of value needs risk decisioning of some type.

So what I actually noticed is a broad trend in the industry, across industry verticals that needed a platform. And specifically, I identified two major gaps across all of these different types of companies, whether it is banks, FinTech or merchants and so on.

The first gap is the fragmentation created by point solutions. And these point solutions only focus on one aspect of the customer life cycle, like there are tools to do identity verification or just credit decisioning, and they actually don't share valuable signals that could help reduce false positives detect sophisticated attacks. So that's the first gap I noticed.

The second gap I noticed across the board was the lack of modern technology like real time data and AI in many solutions, even as fraudsters themselves have started deploying advanced AI tools, like we all know.

So these two gaps were really pervasive across a variety of businesses, which I thought was really interesting as a market trend. And these gaps create significant challenges in effectively reducing risk, ensuring seamless customer experiences.

The core challenge, which just made sense from first principles, even beyond the experience that I actually had, was that there is no 360 view of the user's risk profile that should ideally inform decision making at any point of the customer life cycle. So it's kind of crazy that the risk that you identify during onboarding cannot be shared with the tool that does transaction monitoring, cannot be shared with the tool that does credit decisioning. It just didn't make sense to me, and given that, I noticed that it was really a trend across business of all types.

Nik Milanovic: So when you talk about a 360 degree view or risk profile of the customer, what does that customer look like, and what types of end users do you have in mind when you first started Oscilar?

Neha Narkhede: Oscilar is a platform company, so the way we are approaching the space is really different, in the sense that we're not really creating one kind of point solution, which, in my opinion, is the easy problem to solve. It's the easy product to build, but I don't think it solves the problem, which is that you need to be able to share signals from one point of the customer life cycle to all the other points of the customer life cycle.

So it really made sense that there is a lack of a real unified platform that can assess risk at every point of the customer life cycle. And we've seen companies of all types, and initially we are starting with banks, FinTech FIs but in the future, we will absolutely expand to all the other industry verticals that have online transactions.

We've seen companies significantly reduce false positives, detect fraud earlier, speed up credit approvals by using this unified platform approach. A good percentage of our customers have purchased all our products from onboarding credit, fraud and compliance, and that's happening because of a core reason - that's essentially the core gap in the market and the core gap in technologies.

It's really hard to build a real time data and AI powered platform that can operate risk across the entire customer life cycle. For example, one of our FinTech customers brought all of their siloed solutions under a single decision layer powered by Oscilar. So before the identity verification tool didn't talk to their fraud transaction monitoring system, which obviously creates blind spots.

After unifying everything, they immediately caught 60% more account takeover attacks, slashed manual review times by 85% by leveraging these shared signals, like device fingerprints, behavioral data, which come out of the box with Oscilar. But at the same time, they improved their credit underwriting accuracy by 30% because past fraud checks and identity insights were right there in the same system. Again, something that makes perfect sense when you think about it - it gave them a real time holistic view of the customer, which ultimately led to faster decisions and better risk outcomes, no matter what type of risk it was, onboarding credit, fraud or compliance.

Nik Milanovic: Working in financial services, you hear this a lot about internal systems, that they are fragmented, that you could have a customer who has a product, let's say a customer at a bank that has a mortgage and a savings account and a student loan, and yet each of those products and the relevant customer data live in different silos, such that a relationship manager can't even access the fact that they're a customer of multiple products across the bank. I'm curious if you're kind of unpacking why so many banks and financial institutions operate like this and look at risk systems specifically. Why are they so fragmented at financial services institutions to begin with?

Neha Narkhede: I think risk systems are fragmented largely because many vendors build point solutions that focus on a very narrow slice of the customer journey with its identity verification or AML or credit checks. So these niche tools, the reason for this is, my hypothesis is that they're often actually really easy to stand up quickly, to build quickly, since they make core assumptions about which data points are needed and which data points can you make assumptions around to address a specific type of fraud or risk.

However, each of these point solutions force businesses to send a predefined data set in very rigid formats, leaving very little room for more advanced analysis or the incorporation of additional signals, which is just very key to solving complex types of risk.

So as a result, these institutions end up with really siloed tools that don't communicate with each other or share valuable insights, limiting their ability to get this 360 view of the user's risk profile in real time, no less. What's truly needed is a unified AI driven risk decisioning platform that consolidates risk assessment across every stage of the customer life cycle, not only that, but applies machine learning models specifically that are designed for specific types of risk to accurately predict and mitigate complex risk.

So I think the fragmentation is really caused by vendors who build easy to build point solutions, but they don't quite solve the entire problem, especially in the modern world.

Nik Milanovic: Now, I think this dovetails really well into something that you said at the beginning of this call, which is that Oscilar takes a one platform approach to building a risk management solution. What does it mean to take one platform approach here?

Neha Narkhede: Being a unified one platform in risk management means consolidating all aspects of fraud detection, compliance and decisioning into a single, integrated system that works seamlessly across the entire customer life cycle. At the same time, what we've done really well is we've designed it as a completely modular and flexible platform. So you can easily start with one use case, whatever might be the biggest pain point, and very seamlessly expand to all the other risk use cases. You don't really have to make the hard decision of going all in or not. They actually can very gradually adopt it.

So rather than juggling multiple point solutions, each with its own data format, risk logic, ML models, a unified platform enables you to do a couple things really well:

One is ingest and combine diverse data signals like identity device, behavioral biometrics, transaction histories. You can do all of that in real time, with latency that's sub 100 milliseconds, in many cases, within 15 milliseconds.

The second thing you can do is apply consistent and adaptive decisioning using a shared set of rules, ML models that learn and update across the entire customer life cycle. So this is a big advantage of being able to share signals.

The third thing is providing a 360 view so that any instance, any insights from any interaction inform future decisions, like an onboarding check will inform account monitoring, credit approvals and so on.

And last but not least, you can really streamline operations and governance by eliminating these silos, whether it's across risk or across products like you mentioned, reducing duplication, ensuring uniform compliance processes.

So in a sense, a true unified solution really removes the friction of disparate tools and gives institutions one authoritative source to detect and mitigate fraud, stay compliant and also make smarter decisioning of any type.

Nik Milanovic: When you talk about being able to return values in milliseconds in financial services, that's such an important metric to be able to evaluate yourself by because if you think about financial services interactions, they're usually very quick - being able to pay for something in real time at a payment terminal, needing to be able to verify an account identity to open up a banking app. For instance, it's just critical to be able to process data and then return a solution or a decision quickly. When you think about all the different customer touch points that a bank or FinTech has with their customers, can you walk us through how their end customer uses multiple aspects of the platform together?

Neha Narkhede: Absolutely. I think we've seen, for instance, I can give you a couple examples of the impact of being able to use these multiple products. We've seen customers cut onboarding times by as much as 50% or more after unifying identity checks, fraud detection, AML workflows into a single platform.

So instead of toggling between multiple tools for verification and compliance, everything happens in real time in one place. So you also get this complete 360 view of the user's risk profile that gets updated in a couple milliseconds as every transaction happens and takes place. And that will require sophisticated real time data infrastructure, which, no surprise, Apache Kafka and Confluent power.

On the compliance side, we routinely see an 85% reduction in alert review and case investigation time, thanks to this centralized data and advanced natural language based co-pilot, AI driven alerting, automated documentation, all of these things come together when you look at a unified view.

So not only do new customers get onboarded faster, but compliance teams can actually stay ahead of regulatory requirements without getting this huge operational burden and fraud teams can stay one step ahead of really complex fraud patterns by using this unified view and marrying that with their right AI models.

Nik Milanovic: Would you mind quickly talking me through the different user types or user personas for a product like Oscilar? We've been talking a lot about the end customer, the bank's account holder, but from their perspective, they never actually really interact with Oscilar. And yet, when you talk about the uses of your product, you're talking about fraud teams, you're talking about compliance teams. You're talking about data transparency for the bank to then take and turn around present to its end regulator as well. How do you think about who the user types are that you're serving with Oscilar directly?

Neha Narkhede: That's a great question. I think, given we cover such a broad surface area, it really depends on the type of risk and the product in question.

So for instance, for credit decisioning, we interact with the credit risk analysts which come up with these risk models want to express it very easily without depending on engineering. So that's one of the biggest problems with existing tools, if they're not self service enough. So risk analysts, despite having a tool that apparently does no code decisioning, they actually still have to rely on technical resources.

The other example on the AML side are case reviewers, risk operators that are responsible for flagging ML risks as well as filing SAR reports. So all of that can be done using natural language in the Oscilar platform. So it's truly self service in nature.

And on the fraud side, we actually interact with fraud risk analysts and in many cases, data scientists who can very seamlessly use the platform to create new features, velocity counters, complex features like aggregations, be able to use their ML model, as well as our out of the box ML model, to flag sophisticated types of fraud, whether it is account takeovers or payment fraud,CH fraud, first party fraud. So it really depends on the type of risk problem and the specific product that these banks, FinTech, FIs, are using from Oscilar.

Nik Milanovic: Pivoting a little bit here, thinking about the banking ecosystem in the US, over the last year, we've seen a lot of increased regulatory scrutiny, particularly of sponsor banks. For people who aren't familiar, many FinTech companies in the US, rather than becoming licensed themselves to offer financial products, will work with a sponsor bank, and the bank will effectively sponsor the program as the program manager, and allow the FinTech to acquire customers, but there's been increased exams from Prudential regulators such as the FDIC and OCC who have been really scrutinizing sponsored banks over the past couple years, trying to understand a little bit more about the programs they sponsor and how those programs treat their end users. Now I feel like you're so close to this on a day to day basis. In your opinion, what's been driving this trend of increased scrutiny?

Neha Narkhede: I think regulatory scrutiny of sponsor banks in particular is increasing due to the rapid growth of banking as a service (BaaS) and also the rise of these FinTech bank partnerships. Here's what I think is driving this trend.

There are a couple of things. First is this kind of explosion in FinTech partnerships and compliance gaps. So more fintechs are launching financial products with sponsor banks, but many lack robust compliance programs. So regulators are concerned that some sponsor banks aren't really exercising enough oversight over their FinTech partners' risk and ML programs.

The second thing, I think, is legacy compliance infrastructure is simply not able to keep up. So these traditional compliance systems were not built for the complex oversight required across multiple FinTech partners. That's just a design pattern that they are not built for. So many sponsor banks struggle to monitor compliance across fintechs, across products of different types in real time, which leads to many regulatory gaps, not to mention that they're extremely batch oriented. So there could be days and weeks past before you actually noticed or can raise an alert about some compliance gap.

The third thing I think is happening is high profile enforcement actions and regulatory expectations. Like you mentioned, FDIC, also recent DOJ, OCC, CFPB actions, they highlight the need for stronger controls in these partnerships. And essentially what's happening, I think, is regulators now expect sponsor banks to have very clear visibility into FinTech operations, real time transaction monitoring and too active rather than reactive risk controls.

Last but not least, there's need for a very scalable and unified risk management approach. So sponsoring banks need a centralized platform to oversee multiple FinTech partners, be able to enforce compliance policies across all or a subset of fintechs, depending on which products are in question and also detect risk before it becomes a real regulatory issue.

I think these are the trends, but I've seen our customers, the sponsor banks, the way they are staying ahead is they're adopting a real time compliance monitoring platform that gives them full visibility. Essentially, we call it a command center view, of all the activity across all the fintechs they partner with.

They're also ensuring AML and fraud controls are standardized across all FinTech partners. So tools like Oscilar give you one-click way of saying, "deploy these controls across your FinTech partners that serve a particular kind of product," so you can also segment it intelligently.

And last but not least, I think they're also using AI powered transaction monitoring to detect suspicious patterns early. So regulators aren't just focusing on fintechs anymore. They expect sponsor banks to lead the compliance efforts. So those who proactively strengthen their oversight will not only reduce risk, but they also build sustainable FinTech partnerships in the long run.

Nik Milanovic: I love it. We're really hitting on all the hot button topics here. I'm going to ask in a second about going into specifics on a case study, but two quick questions before that. I have to ask you, you mentioned banking as a service that has definitely been one of the biggest themes in FinTech news over the last year, definitely a mixed year for banking as a service providers. For those listening banking as a service providers are normally platforms that sit between sponsored banks and underlying FinTech programs. What do you think is likely to happen to the banking as a service space?

Neha Narkhede: I think it's interesting. First of all, I don't think these FinTech bank partnerships are going away. I think there is a need for really modern products that can really drive innovation in the space, and whether it's consumer expectations or just availability of the right infrastructure.

I think the right infrastructure is required to enable this innovation, which is introduced because of these FinTech bank partnerships. So I actually think that banking as a service term, or the terminology is gaining really bad traction from brand perspective, but really I think I see it as a broader trend of embedded finance and how we can actually build the right infrastructure that has the right compliance, fraud controls to give them the right technology to enable these kind of partnerships.

So really, I think the broader trend of embedded finance is not going away. It will actually increase the need for the right technology infrastructure to enable these partnerships.

Nik Milanovic: You know, I mentioned kind of two hot button topics. The second one that you brought up was using AI to help banks better monitor and catch fraud in real time. I'm curious how Oscilar is in this AI explosion that it feels like we're seeing fraud feels like a very applicable space, because you have so much unstructured and seemingly uncorrelated data. How is Oscilar leveraging AI in its products?

Neha Narkhede: We leverage AI in a couple of different ways. First, Oscilar creates specific and provides out of the box AI models for specific types of fraud, so they're fine tuned for certain fraud patterns. For example, our ACH fraud model is fine tuned to flag ACH fraud. And in one of our customers called Flus, they were able to increase approval rates by 20% without increasing the fraud rate, and also counter first party fraud.

But we also have specialized ML models for first party fraud, which is really tricky to detect, account takeover fraud, payment fraud of all types and so on. So that's one way in which we leverage AI.

The second thing is our generative AI, pragmatic application of generative AI in the form of a risk co-pilot. So what it's able to do is it can take natural language and create entire workflows and rules, which is really powerful. It really reduces the time it takes to come up with the right strategies, implement them, go live. It also gives you natural language explanations of why something happens - why a decision was made or why a case was created, which really cuts down the manual review time and gives these risk analysts the insight they need in real time to know what is going on, so you can do something about it.

And the third thing you can do is reason, which is, as we all know, one of the powerful features of generative AI is you can actually ask Oscilar AI to give you root cause analysis, and that's something that's coming up very soon. You can ask like, "Hey, why did the fraud rate increase week over week? How much did it increase by? What are the top reasons?" This is really powerful. This has just never been done before. It's a very hard problem.

We've been working on it for close to a year now, experimenting with different models, experimenting with different techniques, to be able to enable this kind of sophistication. So really, you can create models, plot analytical patterns, you can ask the "why" kind of questions, all using Oscilar AI, which I think has, from our customers' feedback, really led to great efficiencies as well as increased capability of detecting risk trends.

Nik Milanovic: Would you mind sharing a specific example? Curious to hear how Oscilar has helped a bank improve their AML Compliance, just walking through one of your case studies.

Neha Narkhede: Last year, we partnered with TransPecos Banks, for instance, which is an innovative Texas based community bank, which is one of the many examples to modernize their AML compliance, specifically in risk management.

So before implementing our AI powered AML platform, they were juggling with multiple silo processes, while also trying to oversee diverse FinTech partners, each with a distinct risk profile. By consolidating AML workflows into one centralized Mission Control, TransPecos Banks gained a real time portfolio view with every FinTech activity.

So they can now automate key tasks like SAR filing, apply advanced ML models for detecting risky transactions, quickly update rules to adapt to new regulations using Oscilar AI. The results were actually really impressive: $3 million in annual cost savings, 40% reduction in AML operations cost, fewer false positives, 70% reduction in alert review and case investigation time, all customer data and risk insights in a single view, and 80% reduction in SAR management time because Oscilar AI can auto fill reports, generate narratives, and provide one-click e-filing.

So this transformation not only streamlined their operations, which is important, but also positioned Trans Pecos to grow its embedded finance offering confidently ensuring compliance that remains at the forefront of that kind of innovation. We're really excited about the partnership with Trans Pecos Banks and many banks of that nature.

Nik Milanovic: No kidding, that's great. I know TransPecos. It's interesting. It's been a mixed year for BaaS providers and sponsored banks, and a lot of sponsored banks have been offloading some of the programs that they work with. TransPecos seems like it's one of the few banks in the last year that's really grown the number of programs it's working with, and I'm sure a big part of that is thanks to platforms like Oscilar that are helping them grow responsibly.

You know one other, I guess you would call it an end user type for a platform like this, is that when you have better visibility into your customers and what they're doing, you also develop a better audit trail, whether you're on the compliance team or the fraud team or the credit risk team, and that audit trail ends up being really important when these banks turn around and have regulatory exams where they need to really be able to show the rationale for decision making, the touch points with the customer and the data on how the customer was actually interacting with the product. And I'm sure that has been a big source of uplift for banks like Trans Pecos, who now really have this full audit record that they can turn and share with their own regulators?

Neha Narkhede: Yeah, absolutely. I think Oscilar actually has a detailed audit trail functionality and also automated model documentation, and that has really helped our banking partners to prove to regulators that they are compliant. And like you said, it's able to track down all the whys behind every decision when required. And we're really excited that our partnership has enabled Trans Pecos Banks to expand their FinTech program so much faster than any other banks in the space.

Nik Milanovic: You know, I'm curious, what are the KPIs, or the success metrics for a customer like Trans Pecos, something like increased onboarding times, better compliance efficiency, lower closing times for cases. I'm curious how your customers think of success when they use Oscilar?

Neha Narkhede: There are a couple of things. One KPI is false positive rate because a lot of the existing compliance tools just have this alert overload, which is one of the biggest problems that exist in the compliance AML space in particular. So this false positive rate, alert overload, alert timing, that's the second - is just how long it takes to investigate cases. That's another KPI.

So we allow them to track in a detailed dashboard all these KPIs, so you know exactly where the inefficiencies are. Also the time it takes to file SAR reports is really time consuming, so that's part of the KPIs as well. And actually, at the end of the day, cost saving - how much efficiency are they able to create, and what cost saving does that translate into? That's, at the end of the day, the most important KPI.

Nik Milanovic: You touched on this a bit, the use of AI in a fraud tooling product like Oscilar. The last couple of years, we've seen this explosion in the space of generative AI specifically, and when you have a new technology, often, some of the first adopters tend to be people who can use it for, let's say, nefarious purposes. Put another way, scammers have been very quick to adopt generative AI specifically for financial scams.

There is a company in Hong Kong that mistakenly sent a $25 million wire to the wrong counterparty because they had a video call with a generative AI deep fake. And so in this kind of arms race where you have bad actors in the system using generative AI, you also have the opportunity for good actors to adopt generative AI. And so I'm curious how you've seen this gen AI growth as a whole, really changing the fraud landscape.

Neha Narkhede: I think recent advancements in generative AI have drastically changed the fraud landscape, making attacks faster, more scalable and harder to detect. There's a couple things that are happening.

First is AI powered fraud at scale. Like you mentioned, fraudsters are using Gen AI to automate large scale attacks, including creating synthetic identities, deep fake audio and video in your example, which is one of the craziest examples, I think, of using generative AI for complex kinds of fraud, automated phishing and social engineering, so these AI generated email chatbots mimicking real executives financial institutions.

The second thing that's happening is traditional fraud detection is just struggling to keep up. So rules based systems are ineffective against these evolving Gen AI powered attacks. As legacy fraud tools rely on these kind of rigid, static data points, like email addresses, then which AI powered fraudsters can easily fabricate. And manual fraud review teams are just overwhelmed as these AI enabled fraud attempts are happening at an unprecedented scale.

The third thing is, you need AI to beat AI, and that's just the only way to keep up with the arms race here. The way it's changing is how AI powered fraud detection is fighting back. To counter these threats, financial institutions need equally advanced AI.

Some examples are real time device and behavioral intentions - detecting AI generated fraud patterns by analyzing device fingerprints, keystroke dynamics, behavioral anomalies and so on. Generative AI risk detection - using ML models trained on emerging fraud tactics to spot synthetic identities, deep fake scams, using multi-pronged strategies, so not just depending on the signals that are shared by the user, but also passive signals, like behavioral biometrics, using data points across the customer journey, profiling the behavior of the user, comparing it to the collective behavior of good users, which is really hard to do from a technology perspective.

This is how we're going to fight back, and is how we're going to use AI to fight AI. That's really, I think the fraud battle is now just the AI versus AI, and only the real time adaptive risk intelligence approach powered by AI can stay ahead of the next generation of fraud threats.

Nik Milanovic: Something that stands out, specifically when you think about different fraud factors and threats, is device intelligence. When Oscilar talks about device intelligence, I'm curious what do you mean? What are you referring to, and why is it so important for a risk assessment?

Neha Narkhede: Device intelligence is becoming essential because it offers real time insights into devices people are using to access these financial services. So for example, by analyzing factors like not just IP addresses, but browser versions, operating systems, but even marrying that with behavioral patterns like what authors are, device and behavior intelligence product does institutions can spot anomalies like sudden device changes or impossible location changes that may signal fraud as more and more transactions move online, device level intelligence really provides a powerful layer of security, which helps distinguish genuine customers from bad actors. That's actually what we've seen in real case studies across our customer base.

Nik Milanovic: Would you be happy to get into, I'd love to be very curious to hear about a couple of different case studies.

Neha Narkhede: A great example of how Oscilar's platform reduces fraud comes from fast growing FinTech, Happy Money, which focuses on responsible lending for over 300,000 members. They needed a solution that could detect sophisticated fraud patterns like synthetic identities without adding friction to the user experience, which I'm sure a lot of modern fintechs appreciate.

So by integrating Oscilar's cognitive intelligence platform, Happy Money was able to do a couple things. One is passively monitor thousands of cognitive signatures during loan applications, as well as ongoing account management to detect these suspicious patterns in real time. We were significantly able to reduce fraud rates by catching synthetic identities and also advanced fraud attempts before they could impact their lending operations.

So this kind of, back to my previous point, marrying broad signals to impact credit risk decisions - that's just really one of the key takeaways of this case study, along with the power of device and behavior signals. And most importantly, they were able to maintain an extremely seamless user experience for legitimate applicants, ensuring that these security enhancements didn't really slow down the onboarding process. So we've been really excited about examples customers like Happy Money, who have really benefited from using our cognitive identity intelligence platform.

Nik Milanovic: Yeah, Happy Money are great, haven't heard their name in a while. Is Oscilar publicly sharing your customer account these days? Or is that heavily guarded state secret?

Neha Narkhede: That's heavily guarded state secret at the moment. I'm sure we will be able to share it in near future.

Nik Milanovic: Fair enough. Well, I'll wait for the press release. I won't leak any kind of non-public information on this podcast. I'm thinking about Oscilar in the context of my own background now, and I was VP of strategy at a credit card issuer called Pedal Card, where, if you're a founder in the FinTech space, especially consumer FinTech, you know this intuitively, new products always get hit with the most fraud immediately. Because they have the lowest level of defenses, and we were no exception.

We built an alternative credit decisioning method that used cash flow data, and it was helpful, but we still had instances of gaming the system and of fraud, just like every other large credit provider. Now you take our example, where we had 400,000 customers, scale that up to card portfolios of a million of 10 million, like the big issuers have. I'm curious, what are the shortcomings in credit decisioning methods, especially traditional credit decisioning methods these days? Why is that still such a blocker for so many consumers?

Neha Narkhede: The traditional credit decisioning methods, in my experience, fall short for many consumers because they rely heavily on static credit bureau data and rigid scoring models that just don't capture the full financial picture. And here's why that's a problem.

First is there's limited data on thin file and no file consumers, many people, especially young adults, immigrants, gig workers, they lack extensive credit history, making them essentially invisible to these traditional scoring models.

There's also inability to incorporate real time financial behavior, so traditional models don't factor in real time income streams, spending patterns or alternative data like rent payments, transaction behaviors, which are just better, higher quality indicators of financial value.

The third thing, third reason why this is a problem like you mentioned, is fraud and identity gap. So relying solely on these credit bureau data points makes institutions vulnerable to synthetic identity, stolen credentials, since fraudsters can manipulate these credit histories.

And last but not least, I think, is this one size fits all approach where traditional models apply the same rigid rules to all applicants failing to adapt to diverse risk profiles of modern consumers. So I think what's needed is a more dynamic, explainable, AI powered approach that leverages real time financial behavior, alternative data sources, adaptive risk models to make smarter and more inclusive credit decisions.

Nik Milanovic: Would you happen to have any specific examples of approval rate improvements or other key metrics that your credit issuing customers have been able to achieve by working with Oscilar?

Neha Narkhede: Yes, we've seen significant improvements in approval rates, fraud reduction and compliance efficiencies also. Some key examples is increased approval rates with predictive models. So a FinTech lender integrated Oscilar's pay prediction ML model and balance prediction ML model, which uses open banking data to assess income stability and future cash flow.

These models are able to predict, for example, the payday for the next 30 days for all types of workers, including gig workers. Balance prediction model can detect fraud banks for the next 30 days. These are really sophisticated models.

So this has enabled the FinTech lender to increase approval rates by up to 20% while maintaining the same risk thresholds, which has really unlocked lending opportunities for gig workers, for them and thin file customers, effectively increasing their total addressable market. So that was really very exciting.

A digital bank was able to reduce fraudulent loan approvals by 70% by leveraging Oscilar's device intelligence behavioral biometric product along with our ML models, which detected synthetic identities and account takeovers in real time.

And last but not least, by unifying fraud detection, credit decisions, compliance into a single AI powered platform, one of our customers reduced loan approval times from minutes to literally seconds, which resulted in the fastest onboarding time and a much better, elevated customer experience.

So this really, I think, highlights how real time, explainable AI alternative data, unified decisions really affect and result in both higher approvals as well as stronger fraud detection, which is ultimately what helps these businesses scale effectively and safely.

Nik Milanovic: You know, it just seems like there's so much low hanging fruit in this space. It's wild to me that until recently, there weren't products available for reporting rent payments to the credit bureaus. You know, you have recurring payments that you make as a consumer, and they're just not or weren't factored at all into your credit worthiness.

And so you really need companies that can help you process, I guess what you'd call alternative data, but really is relevant data about consumer credit worthiness in order to be able to effectively underwrite them. You know, we spend so much time in FinTech talking about unbanked consumers, under banked consumers, but credit visibility is as pernicious a problem for consumers in the US.

And the uplift that you've been able to help these credit issuers experience, I'm sure, has resulted in a lot of happy end customers who are probably turned down for credit elsewhere.

Have you got five more minutes? You are on the road and so always difficult to catch a full hour of time, but I appreciate everything that you share with us about the product today. Maybe broadening the lens a little bit, I'd be curious where you see risk management heading in the next three to five years more broadly for the industry.

Neha Narkhede: Risk management is evolving very rapidly, which has been exciting and rewarding at the same time, but over the next three to five years I think we should all expect to see a couple of major shifts.

First is hyper personalized real time risk decisions. So traditional risk models will give way to real time AI driven decisions that dynamically adapt to each customer's risk profile. So instead of static rules, just static rules, financial institutions will use live, behavioral financial data to assess risk at every touch point, from onboarding to trade decisions to fraud detection.

The second trend, I think is happening, and will happen even more rapidly, is AI versus AI, the battle against AI powered fraud. So as fraudsters increasingly leverage generative AI to create synthetic identities, deep fake scams, financial institutions will need even more advanced AI to detect and counter these attacks.

In my experience, that is a really smart combination of traditional predictive machine learning models as well as generative AI models, which we found to be really effective methodology in this battle against AI powered fraud.

And last but not least, on the compliance front, unified compliance, embedded risk platforms. So the era of these fragmented risk solutions is coming to an end. We are moving to an era of consolidation. Institutions will move towards unified platforms that seamlessly integrate fraud, AML, credit risk decisioning into a single system, and this will allow better cross functional oversight, automated regulatory compliance, even more efficient risk management without increasing operational costs.

Ultimately, I think risk management is shifting from being a defensive function to a strategic enabler. It's going to help businesses grow faster, safer, more inclusively, while staying ahead of evolving threats. And as consumers, we're going to appreciate better safeguards and much better customer experiences, which is just a win win for everyone.

Nik Milanovic: So how should financial institutions be thinking now if they're looking three to five years down the road, and want to be able to better prepare for emerging threats?

Neha Narkhede: Any financial institution should take a proactive and AI driven approach to risk management to really be able to stay ahead of emerging threats, which is responding to sort of historical threats that they might have seen more commonly in the past.

I think what they should be doing now is adopting real time AI powered decisions. Fraudsters are using AI and automation to scale attacks. So relying on static rules outdated models is just no longer enough. So you need a combination of data sources. You need ML Model Driven risk scoring to detect these evolving threats instantly. You need real time feedback loops to retrain your ML models, which is tough from a technology perspective, we're definitely seeing it's possible, as we've seen at Oscilar.

The second thing that these institutions are going to need to do is break down silos. Many banks still rely on fragmented, disconnected tools. A unified platform is going to give them the oversight and real time collaboration across different risk functions and enhanced AI explainability and regulatory readiness, like we talked about. Regulators are scrutinizing AI decision making, financial institutions must ensure that the risk models are explainable, auditable, compliant with evolving standards, and just be future proof with adaptive infrastructure.

So this risk landscape is just moving too fast for static systems. Institutions will need flexible AI, API driven platforms that can quickly adapt to new threats, integrate emerging data sources, continuously refine these models in real time.

So by taking these steps now and just accepting these trends, financial institutions can stay ahead of emerging threats, reducing fraud, improving compliance efficiency, and I think just building a safer, more resilient financial ecosystem at the end of the day.

Nik Milanovic: Neha, thank you again so much for making the time to talk. It's really great to have you at This Week in Fintech. You know, all these conversations go in different directions. I feel like sometimes we talk about personal life outside of work, sometimes we talk about company history. But I'm glad that we were able to really just dig down into the specifics in this space that I think many, even in financial services, do not understand at a deep level.

This week, our CEO & Co-founder, Neha Narkhede, joined Nik Milanović on This Week in Fintech for an in-depth discussion about the future of financial risk management and Oscilar's innovative approach to AI-powered risk decisioning.

Key Highlights

In this conversation, Neha explains how a unified platform approach to AI Risk Decisioning creates a comprehensive view of fraud, credit, and compliance risks. She discusses:

  • The critical gaps in today's fragmented risk management landscape and why point solutions are insufficient for modern financial institutions

  • How Oscilar's unified platform enables real-time risk decisions by combining fraud detection, compliance monitoring, and credit decisioning

  • The transformative impact of shared risk signals across the customer lifecycle

  • Real customer success stories, including how Trans Pecos Banks achieved 40% reduction in AML operations costs and 70% reduction in alert review time

  • The growing importance of AI in combating sophisticated fraud techniques, from synthetic identities to deep fakes

  • The increasing regulatory scrutiny of sponsor banks and why real-time compliance monitoring is crucial

  • Her vision for the future of risk management, including hyper-personalized decisioning and unified compliance platforms

For risk operations professionals, compliance teams, and fintech leaders, this conversation provides valuable insights into building more resilient financial systems through unified, AI-powered risk management.

You can listen to the full episode here

Full Episode Transcript

Nik Milanovic: Thank you everyone for joining us today. We are back for another This Week in Fintech podcast, and I am very excited for our guest today. Our guest is a founder, a repeat founder, I should say, who has been in the financial services and fintech space for a while, and I'm very excited to hear from her. Her name is Neha Narkhede, and she is the founder of Oscilar.

For those of you who may not yet be familiar with Oscilar, Oscilar is an AI risk decisioning platform that helps organizations manage onboarding, fraud, credit and compliance risks.

Neha is a repeat founder who spent a while within the space building out new products, and we're going to dig in today to what Oscilar is building, what their vision is for the future, a few of their customer case studies and what comes next.

Thanks again for joining us and Neha, thank you so much for making the time to be with us.

Neha Narkhede: Thank you, Nik. Very excited to be here, just kicking off. I know we've done a few events together This Week in Fintech and Oscilar. We've partnered in the UK. We've partnered in the US. We've partnered in Brazil. It feels like you're everywhere as Oscilar is growing around the world. Oscilar is a global company, and we're a fast growing business that operates in North America, Europe, Latin and also in the Middle East.

Nik Milanovic: For those who don't know, Neha previously founded a company called Confluent, where I believe you were there for ten years as their CTO chief product officer, and then later on a board member, building a company for real time data and Apache Kafka. I'm curious, could you walk us through your journey from founding Confluent in Kafka to founding Oscilar?

Neha Narkhede: It's been a really fun journey at Confluent. I co-founded Confluent to build a real time data streaming platform powered by a very popular open source system called Apache Kafka, which I was one of the original creators for, and we help businesses all around the world harness real time data and AI to transform their operations.

While working with the leading banks, fintechs, merchants, travel and hospitality, I saw firsthand the fraud and risk management were among the most critical and high impact use cases of real time data and AI in the modern world. However, the existing tools for agile, real time, fraud prevention, risk and compliance, they fell short in a number of ways, and people were trying to work around it and trying to build various tools around it, which actually didn't make a lot of sense.

People are seeing more and more transactions of value moving online at a very fast pace since the COVID pandemic, and they still are without modern and sophisticated risk infrastructure. So recognizing this massive trend and really driven by my passion to secure online transactions, I co-founded Oscilar to empower businesses to detect fraud, manage compliance, make better credit risk decisioning with cutting edge data analysis and machine learning all through one unified risk platform, which is being done for the very first time.

Nik Milanovic: And I noticed that when you founded and announced Oscilar, something that was interesting to me was that you and your co-founder, Sasha Kulkarni also self-funded it to get it off the ground, which to me, stands out as clear example of high conviction in the product you're building, that this is something that you're willing to self fund just to get up and off the ground.

You mentioned while you were at Confluent, you were working with a set of risk tools - tools that you just didn't feel really met your needs, and so it kind of started this burning desire to actually build a better risk product based on your own experience. I'm curious, what were the moments at Confluent that made you realize that there was a real gap in the market and opportunity for building a tool like this? What moments gave you such high conviction that you were willing to start over from the ground up as a repeat entrepreneur and self fund to get this product off the ground and up and running?

Neha Narkhede: Given how big the company is, I saw companies across different industry verticals struggle with the same problems. And the trend I saw is, every industry vertical nearly has some online transactions of some type, and any online transaction of value needs risk decisioning of some type.

So what I actually noticed is a broad trend in the industry, across industry verticals that needed a platform. And specifically, I identified two major gaps across all of these different types of companies, whether it is banks, FinTech or merchants and so on.

The first gap is the fragmentation created by point solutions. And these point solutions only focus on one aspect of the customer life cycle, like there are tools to do identity verification or just credit decisioning, and they actually don't share valuable signals that could help reduce false positives detect sophisticated attacks. So that's the first gap I noticed.

The second gap I noticed across the board was the lack of modern technology like real time data and AI in many solutions, even as fraudsters themselves have started deploying advanced AI tools, like we all know.

So these two gaps were really pervasive across a variety of businesses, which I thought was really interesting as a market trend. And these gaps create significant challenges in effectively reducing risk, ensuring seamless customer experiences.

The core challenge, which just made sense from first principles, even beyond the experience that I actually had, was that there is no 360 view of the user's risk profile that should ideally inform decision making at any point of the customer life cycle. So it's kind of crazy that the risk that you identify during onboarding cannot be shared with the tool that does transaction monitoring, cannot be shared with the tool that does credit decisioning. It just didn't make sense to me, and given that, I noticed that it was really a trend across business of all types.

Nik Milanovic: So when you talk about a 360 degree view or risk profile of the customer, what does that customer look like, and what types of end users do you have in mind when you first started Oscilar?

Neha Narkhede: Oscilar is a platform company, so the way we are approaching the space is really different, in the sense that we're not really creating one kind of point solution, which, in my opinion, is the easy problem to solve. It's the easy product to build, but I don't think it solves the problem, which is that you need to be able to share signals from one point of the customer life cycle to all the other points of the customer life cycle.

So it really made sense that there is a lack of a real unified platform that can assess risk at every point of the customer life cycle. And we've seen companies of all types, and initially we are starting with banks, FinTech FIs but in the future, we will absolutely expand to all the other industry verticals that have online transactions.

We've seen companies significantly reduce false positives, detect fraud earlier, speed up credit approvals by using this unified platform approach. A good percentage of our customers have purchased all our products from onboarding credit, fraud and compliance, and that's happening because of a core reason - that's essentially the core gap in the market and the core gap in technologies.

It's really hard to build a real time data and AI powered platform that can operate risk across the entire customer life cycle. For example, one of our FinTech customers brought all of their siloed solutions under a single decision layer powered by Oscilar. So before the identity verification tool didn't talk to their fraud transaction monitoring system, which obviously creates blind spots.

After unifying everything, they immediately caught 60% more account takeover attacks, slashed manual review times by 85% by leveraging these shared signals, like device fingerprints, behavioral data, which come out of the box with Oscilar. But at the same time, they improved their credit underwriting accuracy by 30% because past fraud checks and identity insights were right there in the same system. Again, something that makes perfect sense when you think about it - it gave them a real time holistic view of the customer, which ultimately led to faster decisions and better risk outcomes, no matter what type of risk it was, onboarding credit, fraud or compliance.

Nik Milanovic: Working in financial services, you hear this a lot about internal systems, that they are fragmented, that you could have a customer who has a product, let's say a customer at a bank that has a mortgage and a savings account and a student loan, and yet each of those products and the relevant customer data live in different silos, such that a relationship manager can't even access the fact that they're a customer of multiple products across the bank. I'm curious if you're kind of unpacking why so many banks and financial institutions operate like this and look at risk systems specifically. Why are they so fragmented at financial services institutions to begin with?

Neha Narkhede: I think risk systems are fragmented largely because many vendors build point solutions that focus on a very narrow slice of the customer journey with its identity verification or AML or credit checks. So these niche tools, the reason for this is, my hypothesis is that they're often actually really easy to stand up quickly, to build quickly, since they make core assumptions about which data points are needed and which data points can you make assumptions around to address a specific type of fraud or risk.

However, each of these point solutions force businesses to send a predefined data set in very rigid formats, leaving very little room for more advanced analysis or the incorporation of additional signals, which is just very key to solving complex types of risk.

So as a result, these institutions end up with really siloed tools that don't communicate with each other or share valuable insights, limiting their ability to get this 360 view of the user's risk profile in real time, no less. What's truly needed is a unified AI driven risk decisioning platform that consolidates risk assessment across every stage of the customer life cycle, not only that, but applies machine learning models specifically that are designed for specific types of risk to accurately predict and mitigate complex risk.

So I think the fragmentation is really caused by vendors who build easy to build point solutions, but they don't quite solve the entire problem, especially in the modern world.

Nik Milanovic: Now, I think this dovetails really well into something that you said at the beginning of this call, which is that Oscilar takes a one platform approach to building a risk management solution. What does it mean to take one platform approach here?

Neha Narkhede: Being a unified one platform in risk management means consolidating all aspects of fraud detection, compliance and decisioning into a single, integrated system that works seamlessly across the entire customer life cycle. At the same time, what we've done really well is we've designed it as a completely modular and flexible platform. So you can easily start with one use case, whatever might be the biggest pain point, and very seamlessly expand to all the other risk use cases. You don't really have to make the hard decision of going all in or not. They actually can very gradually adopt it.

So rather than juggling multiple point solutions, each with its own data format, risk logic, ML models, a unified platform enables you to do a couple things really well:

One is ingest and combine diverse data signals like identity device, behavioral biometrics, transaction histories. You can do all of that in real time, with latency that's sub 100 milliseconds, in many cases, within 15 milliseconds.

The second thing you can do is apply consistent and adaptive decisioning using a shared set of rules, ML models that learn and update across the entire customer life cycle. So this is a big advantage of being able to share signals.

The third thing is providing a 360 view so that any instance, any insights from any interaction inform future decisions, like an onboarding check will inform account monitoring, credit approvals and so on.

And last but not least, you can really streamline operations and governance by eliminating these silos, whether it's across risk or across products like you mentioned, reducing duplication, ensuring uniform compliance processes.

So in a sense, a true unified solution really removes the friction of disparate tools and gives institutions one authoritative source to detect and mitigate fraud, stay compliant and also make smarter decisioning of any type.

Nik Milanovic: When you talk about being able to return values in milliseconds in financial services, that's such an important metric to be able to evaluate yourself by because if you think about financial services interactions, they're usually very quick - being able to pay for something in real time at a payment terminal, needing to be able to verify an account identity to open up a banking app. For instance, it's just critical to be able to process data and then return a solution or a decision quickly. When you think about all the different customer touch points that a bank or FinTech has with their customers, can you walk us through how their end customer uses multiple aspects of the platform together?

Neha Narkhede: Absolutely. I think we've seen, for instance, I can give you a couple examples of the impact of being able to use these multiple products. We've seen customers cut onboarding times by as much as 50% or more after unifying identity checks, fraud detection, AML workflows into a single platform.

So instead of toggling between multiple tools for verification and compliance, everything happens in real time in one place. So you also get this complete 360 view of the user's risk profile that gets updated in a couple milliseconds as every transaction happens and takes place. And that will require sophisticated real time data infrastructure, which, no surprise, Apache Kafka and Confluent power.

On the compliance side, we routinely see an 85% reduction in alert review and case investigation time, thanks to this centralized data and advanced natural language based co-pilot, AI driven alerting, automated documentation, all of these things come together when you look at a unified view.

So not only do new customers get onboarded faster, but compliance teams can actually stay ahead of regulatory requirements without getting this huge operational burden and fraud teams can stay one step ahead of really complex fraud patterns by using this unified view and marrying that with their right AI models.

Nik Milanovic: Would you mind quickly talking me through the different user types or user personas for a product like Oscilar? We've been talking a lot about the end customer, the bank's account holder, but from their perspective, they never actually really interact with Oscilar. And yet, when you talk about the uses of your product, you're talking about fraud teams, you're talking about compliance teams. You're talking about data transparency for the bank to then take and turn around present to its end regulator as well. How do you think about who the user types are that you're serving with Oscilar directly?

Neha Narkhede: That's a great question. I think, given we cover such a broad surface area, it really depends on the type of risk and the product in question.

So for instance, for credit decisioning, we interact with the credit risk analysts which come up with these risk models want to express it very easily without depending on engineering. So that's one of the biggest problems with existing tools, if they're not self service enough. So risk analysts, despite having a tool that apparently does no code decisioning, they actually still have to rely on technical resources.

The other example on the AML side are case reviewers, risk operators that are responsible for flagging ML risks as well as filing SAR reports. So all of that can be done using natural language in the Oscilar platform. So it's truly self service in nature.

And on the fraud side, we actually interact with fraud risk analysts and in many cases, data scientists who can very seamlessly use the platform to create new features, velocity counters, complex features like aggregations, be able to use their ML model, as well as our out of the box ML model, to flag sophisticated types of fraud, whether it is account takeovers or payment fraud,CH fraud, first party fraud. So it really depends on the type of risk problem and the specific product that these banks, FinTech, FIs, are using from Oscilar.

Nik Milanovic: Pivoting a little bit here, thinking about the banking ecosystem in the US, over the last year, we've seen a lot of increased regulatory scrutiny, particularly of sponsor banks. For people who aren't familiar, many FinTech companies in the US, rather than becoming licensed themselves to offer financial products, will work with a sponsor bank, and the bank will effectively sponsor the program as the program manager, and allow the FinTech to acquire customers, but there's been increased exams from Prudential regulators such as the FDIC and OCC who have been really scrutinizing sponsored banks over the past couple years, trying to understand a little bit more about the programs they sponsor and how those programs treat their end users. Now I feel like you're so close to this on a day to day basis. In your opinion, what's been driving this trend of increased scrutiny?

Neha Narkhede: I think regulatory scrutiny of sponsor banks in particular is increasing due to the rapid growth of banking as a service (BaaS) and also the rise of these FinTech bank partnerships. Here's what I think is driving this trend.

There are a couple of things. First is this kind of explosion in FinTech partnerships and compliance gaps. So more fintechs are launching financial products with sponsor banks, but many lack robust compliance programs. So regulators are concerned that some sponsor banks aren't really exercising enough oversight over their FinTech partners' risk and ML programs.

The second thing, I think, is legacy compliance infrastructure is simply not able to keep up. So these traditional compliance systems were not built for the complex oversight required across multiple FinTech partners. That's just a design pattern that they are not built for. So many sponsor banks struggle to monitor compliance across fintechs, across products of different types in real time, which leads to many regulatory gaps, not to mention that they're extremely batch oriented. So there could be days and weeks past before you actually noticed or can raise an alert about some compliance gap.

The third thing I think is happening is high profile enforcement actions and regulatory expectations. Like you mentioned, FDIC, also recent DOJ, OCC, CFPB actions, they highlight the need for stronger controls in these partnerships. And essentially what's happening, I think, is regulators now expect sponsor banks to have very clear visibility into FinTech operations, real time transaction monitoring and too active rather than reactive risk controls.

Last but not least, there's need for a very scalable and unified risk management approach. So sponsoring banks need a centralized platform to oversee multiple FinTech partners, be able to enforce compliance policies across all or a subset of fintechs, depending on which products are in question and also detect risk before it becomes a real regulatory issue.

I think these are the trends, but I've seen our customers, the sponsor banks, the way they are staying ahead is they're adopting a real time compliance monitoring platform that gives them full visibility. Essentially, we call it a command center view, of all the activity across all the fintechs they partner with.

They're also ensuring AML and fraud controls are standardized across all FinTech partners. So tools like Oscilar give you one-click way of saying, "deploy these controls across your FinTech partners that serve a particular kind of product," so you can also segment it intelligently.

And last but not least, I think they're also using AI powered transaction monitoring to detect suspicious patterns early. So regulators aren't just focusing on fintechs anymore. They expect sponsor banks to lead the compliance efforts. So those who proactively strengthen their oversight will not only reduce risk, but they also build sustainable FinTech partnerships in the long run.

Nik Milanovic: I love it. We're really hitting on all the hot button topics here. I'm going to ask in a second about going into specifics on a case study, but two quick questions before that. I have to ask you, you mentioned banking as a service that has definitely been one of the biggest themes in FinTech news over the last year, definitely a mixed year for banking as a service providers. For those listening banking as a service providers are normally platforms that sit between sponsored banks and underlying FinTech programs. What do you think is likely to happen to the banking as a service space?

Neha Narkhede: I think it's interesting. First of all, I don't think these FinTech bank partnerships are going away. I think there is a need for really modern products that can really drive innovation in the space, and whether it's consumer expectations or just availability of the right infrastructure.

I think the right infrastructure is required to enable this innovation, which is introduced because of these FinTech bank partnerships. So I actually think that banking as a service term, or the terminology is gaining really bad traction from brand perspective, but really I think I see it as a broader trend of embedded finance and how we can actually build the right infrastructure that has the right compliance, fraud controls to give them the right technology to enable these kind of partnerships.

So really, I think the broader trend of embedded finance is not going away. It will actually increase the need for the right technology infrastructure to enable these partnerships.

Nik Milanovic: You know, I mentioned kind of two hot button topics. The second one that you brought up was using AI to help banks better monitor and catch fraud in real time. I'm curious how Oscilar is in this AI explosion that it feels like we're seeing fraud feels like a very applicable space, because you have so much unstructured and seemingly uncorrelated data. How is Oscilar leveraging AI in its products?

Neha Narkhede: We leverage AI in a couple of different ways. First, Oscilar creates specific and provides out of the box AI models for specific types of fraud, so they're fine tuned for certain fraud patterns. For example, our ACH fraud model is fine tuned to flag ACH fraud. And in one of our customers called Flus, they were able to increase approval rates by 20% without increasing the fraud rate, and also counter first party fraud.

But we also have specialized ML models for first party fraud, which is really tricky to detect, account takeover fraud, payment fraud of all types and so on. So that's one way in which we leverage AI.

The second thing is our generative AI, pragmatic application of generative AI in the form of a risk co-pilot. So what it's able to do is it can take natural language and create entire workflows and rules, which is really powerful. It really reduces the time it takes to come up with the right strategies, implement them, go live. It also gives you natural language explanations of why something happens - why a decision was made or why a case was created, which really cuts down the manual review time and gives these risk analysts the insight they need in real time to know what is going on, so you can do something about it.

And the third thing you can do is reason, which is, as we all know, one of the powerful features of generative AI is you can actually ask Oscilar AI to give you root cause analysis, and that's something that's coming up very soon. You can ask like, "Hey, why did the fraud rate increase week over week? How much did it increase by? What are the top reasons?" This is really powerful. This has just never been done before. It's a very hard problem.

We've been working on it for close to a year now, experimenting with different models, experimenting with different techniques, to be able to enable this kind of sophistication. So really, you can create models, plot analytical patterns, you can ask the "why" kind of questions, all using Oscilar AI, which I think has, from our customers' feedback, really led to great efficiencies as well as increased capability of detecting risk trends.

Nik Milanovic: Would you mind sharing a specific example? Curious to hear how Oscilar has helped a bank improve their AML Compliance, just walking through one of your case studies.

Neha Narkhede: Last year, we partnered with TransPecos Banks, for instance, which is an innovative Texas based community bank, which is one of the many examples to modernize their AML compliance, specifically in risk management.

So before implementing our AI powered AML platform, they were juggling with multiple silo processes, while also trying to oversee diverse FinTech partners, each with a distinct risk profile. By consolidating AML workflows into one centralized Mission Control, TransPecos Banks gained a real time portfolio view with every FinTech activity.

So they can now automate key tasks like SAR filing, apply advanced ML models for detecting risky transactions, quickly update rules to adapt to new regulations using Oscilar AI. The results were actually really impressive: $3 million in annual cost savings, 40% reduction in AML operations cost, fewer false positives, 70% reduction in alert review and case investigation time, all customer data and risk insights in a single view, and 80% reduction in SAR management time because Oscilar AI can auto fill reports, generate narratives, and provide one-click e-filing.

So this transformation not only streamlined their operations, which is important, but also positioned Trans Pecos to grow its embedded finance offering confidently ensuring compliance that remains at the forefront of that kind of innovation. We're really excited about the partnership with Trans Pecos Banks and many banks of that nature.

Nik Milanovic: No kidding, that's great. I know TransPecos. It's interesting. It's been a mixed year for BaaS providers and sponsored banks, and a lot of sponsored banks have been offloading some of the programs that they work with. TransPecos seems like it's one of the few banks in the last year that's really grown the number of programs it's working with, and I'm sure a big part of that is thanks to platforms like Oscilar that are helping them grow responsibly.

You know one other, I guess you would call it an end user type for a platform like this, is that when you have better visibility into your customers and what they're doing, you also develop a better audit trail, whether you're on the compliance team or the fraud team or the credit risk team, and that audit trail ends up being really important when these banks turn around and have regulatory exams where they need to really be able to show the rationale for decision making, the touch points with the customer and the data on how the customer was actually interacting with the product. And I'm sure that has been a big source of uplift for banks like Trans Pecos, who now really have this full audit record that they can turn and share with their own regulators?

Neha Narkhede: Yeah, absolutely. I think Oscilar actually has a detailed audit trail functionality and also automated model documentation, and that has really helped our banking partners to prove to regulators that they are compliant. And like you said, it's able to track down all the whys behind every decision when required. And we're really excited that our partnership has enabled Trans Pecos Banks to expand their FinTech program so much faster than any other banks in the space.

Nik Milanovic: You know, I'm curious, what are the KPIs, or the success metrics for a customer like Trans Pecos, something like increased onboarding times, better compliance efficiency, lower closing times for cases. I'm curious how your customers think of success when they use Oscilar?

Neha Narkhede: There are a couple of things. One KPI is false positive rate because a lot of the existing compliance tools just have this alert overload, which is one of the biggest problems that exist in the compliance AML space in particular. So this false positive rate, alert overload, alert timing, that's the second - is just how long it takes to investigate cases. That's another KPI.

So we allow them to track in a detailed dashboard all these KPIs, so you know exactly where the inefficiencies are. Also the time it takes to file SAR reports is really time consuming, so that's part of the KPIs as well. And actually, at the end of the day, cost saving - how much efficiency are they able to create, and what cost saving does that translate into? That's, at the end of the day, the most important KPI.

Nik Milanovic: You touched on this a bit, the use of AI in a fraud tooling product like Oscilar. The last couple of years, we've seen this explosion in the space of generative AI specifically, and when you have a new technology, often, some of the first adopters tend to be people who can use it for, let's say, nefarious purposes. Put another way, scammers have been very quick to adopt generative AI specifically for financial scams.

There is a company in Hong Kong that mistakenly sent a $25 million wire to the wrong counterparty because they had a video call with a generative AI deep fake. And so in this kind of arms race where you have bad actors in the system using generative AI, you also have the opportunity for good actors to adopt generative AI. And so I'm curious how you've seen this gen AI growth as a whole, really changing the fraud landscape.

Neha Narkhede: I think recent advancements in generative AI have drastically changed the fraud landscape, making attacks faster, more scalable and harder to detect. There's a couple things that are happening.

First is AI powered fraud at scale. Like you mentioned, fraudsters are using Gen AI to automate large scale attacks, including creating synthetic identities, deep fake audio and video in your example, which is one of the craziest examples, I think, of using generative AI for complex kinds of fraud, automated phishing and social engineering, so these AI generated email chatbots mimicking real executives financial institutions.

The second thing that's happening is traditional fraud detection is just struggling to keep up. So rules based systems are ineffective against these evolving Gen AI powered attacks. As legacy fraud tools rely on these kind of rigid, static data points, like email addresses, then which AI powered fraudsters can easily fabricate. And manual fraud review teams are just overwhelmed as these AI enabled fraud attempts are happening at an unprecedented scale.

The third thing is, you need AI to beat AI, and that's just the only way to keep up with the arms race here. The way it's changing is how AI powered fraud detection is fighting back. To counter these threats, financial institutions need equally advanced AI.

Some examples are real time device and behavioral intentions - detecting AI generated fraud patterns by analyzing device fingerprints, keystroke dynamics, behavioral anomalies and so on. Generative AI risk detection - using ML models trained on emerging fraud tactics to spot synthetic identities, deep fake scams, using multi-pronged strategies, so not just depending on the signals that are shared by the user, but also passive signals, like behavioral biometrics, using data points across the customer journey, profiling the behavior of the user, comparing it to the collective behavior of good users, which is really hard to do from a technology perspective.

This is how we're going to fight back, and is how we're going to use AI to fight AI. That's really, I think the fraud battle is now just the AI versus AI, and only the real time adaptive risk intelligence approach powered by AI can stay ahead of the next generation of fraud threats.

Nik Milanovic: Something that stands out, specifically when you think about different fraud factors and threats, is device intelligence. When Oscilar talks about device intelligence, I'm curious what do you mean? What are you referring to, and why is it so important for a risk assessment?

Neha Narkhede: Device intelligence is becoming essential because it offers real time insights into devices people are using to access these financial services. So for example, by analyzing factors like not just IP addresses, but browser versions, operating systems, but even marrying that with behavioral patterns like what authors are, device and behavior intelligence product does institutions can spot anomalies like sudden device changes or impossible location changes that may signal fraud as more and more transactions move online, device level intelligence really provides a powerful layer of security, which helps distinguish genuine customers from bad actors. That's actually what we've seen in real case studies across our customer base.

Nik Milanovic: Would you be happy to get into, I'd love to be very curious to hear about a couple of different case studies.

Neha Narkhede: A great example of how Oscilar's platform reduces fraud comes from fast growing FinTech, Happy Money, which focuses on responsible lending for over 300,000 members. They needed a solution that could detect sophisticated fraud patterns like synthetic identities without adding friction to the user experience, which I'm sure a lot of modern fintechs appreciate.

So by integrating Oscilar's cognitive intelligence platform, Happy Money was able to do a couple things. One is passively monitor thousands of cognitive signatures during loan applications, as well as ongoing account management to detect these suspicious patterns in real time. We were significantly able to reduce fraud rates by catching synthetic identities and also advanced fraud attempts before they could impact their lending operations.

So this kind of, back to my previous point, marrying broad signals to impact credit risk decisions - that's just really one of the key takeaways of this case study, along with the power of device and behavior signals. And most importantly, they were able to maintain an extremely seamless user experience for legitimate applicants, ensuring that these security enhancements didn't really slow down the onboarding process. So we've been really excited about examples customers like Happy Money, who have really benefited from using our cognitive identity intelligence platform.

Nik Milanovic: Yeah, Happy Money are great, haven't heard their name in a while. Is Oscilar publicly sharing your customer account these days? Or is that heavily guarded state secret?

Neha Narkhede: That's heavily guarded state secret at the moment. I'm sure we will be able to share it in near future.

Nik Milanovic: Fair enough. Well, I'll wait for the press release. I won't leak any kind of non-public information on this podcast. I'm thinking about Oscilar in the context of my own background now, and I was VP of strategy at a credit card issuer called Pedal Card, where, if you're a founder in the FinTech space, especially consumer FinTech, you know this intuitively, new products always get hit with the most fraud immediately. Because they have the lowest level of defenses, and we were no exception.

We built an alternative credit decisioning method that used cash flow data, and it was helpful, but we still had instances of gaming the system and of fraud, just like every other large credit provider. Now you take our example, where we had 400,000 customers, scale that up to card portfolios of a million of 10 million, like the big issuers have. I'm curious, what are the shortcomings in credit decisioning methods, especially traditional credit decisioning methods these days? Why is that still such a blocker for so many consumers?

Neha Narkhede: The traditional credit decisioning methods, in my experience, fall short for many consumers because they rely heavily on static credit bureau data and rigid scoring models that just don't capture the full financial picture. And here's why that's a problem.

First is there's limited data on thin file and no file consumers, many people, especially young adults, immigrants, gig workers, they lack extensive credit history, making them essentially invisible to these traditional scoring models.

There's also inability to incorporate real time financial behavior, so traditional models don't factor in real time income streams, spending patterns or alternative data like rent payments, transaction behaviors, which are just better, higher quality indicators of financial value.

The third thing, third reason why this is a problem like you mentioned, is fraud and identity gap. So relying solely on these credit bureau data points makes institutions vulnerable to synthetic identity, stolen credentials, since fraudsters can manipulate these credit histories.

And last but not least, I think, is this one size fits all approach where traditional models apply the same rigid rules to all applicants failing to adapt to diverse risk profiles of modern consumers. So I think what's needed is a more dynamic, explainable, AI powered approach that leverages real time financial behavior, alternative data sources, adaptive risk models to make smarter and more inclusive credit decisions.

Nik Milanovic: Would you happen to have any specific examples of approval rate improvements or other key metrics that your credit issuing customers have been able to achieve by working with Oscilar?

Neha Narkhede: Yes, we've seen significant improvements in approval rates, fraud reduction and compliance efficiencies also. Some key examples is increased approval rates with predictive models. So a FinTech lender integrated Oscilar's pay prediction ML model and balance prediction ML model, which uses open banking data to assess income stability and future cash flow.

These models are able to predict, for example, the payday for the next 30 days for all types of workers, including gig workers. Balance prediction model can detect fraud banks for the next 30 days. These are really sophisticated models.

So this has enabled the FinTech lender to increase approval rates by up to 20% while maintaining the same risk thresholds, which has really unlocked lending opportunities for gig workers, for them and thin file customers, effectively increasing their total addressable market. So that was really very exciting.

A digital bank was able to reduce fraudulent loan approvals by 70% by leveraging Oscilar's device intelligence behavioral biometric product along with our ML models, which detected synthetic identities and account takeovers in real time.

And last but not least, by unifying fraud detection, credit decisions, compliance into a single AI powered platform, one of our customers reduced loan approval times from minutes to literally seconds, which resulted in the fastest onboarding time and a much better, elevated customer experience.

So this really, I think, highlights how real time, explainable AI alternative data, unified decisions really affect and result in both higher approvals as well as stronger fraud detection, which is ultimately what helps these businesses scale effectively and safely.

Nik Milanovic: You know, it just seems like there's so much low hanging fruit in this space. It's wild to me that until recently, there weren't products available for reporting rent payments to the credit bureaus. You know, you have recurring payments that you make as a consumer, and they're just not or weren't factored at all into your credit worthiness.

And so you really need companies that can help you process, I guess what you'd call alternative data, but really is relevant data about consumer credit worthiness in order to be able to effectively underwrite them. You know, we spend so much time in FinTech talking about unbanked consumers, under banked consumers, but credit visibility is as pernicious a problem for consumers in the US.

And the uplift that you've been able to help these credit issuers experience, I'm sure, has resulted in a lot of happy end customers who are probably turned down for credit elsewhere.

Have you got five more minutes? You are on the road and so always difficult to catch a full hour of time, but I appreciate everything that you share with us about the product today. Maybe broadening the lens a little bit, I'd be curious where you see risk management heading in the next three to five years more broadly for the industry.

Neha Narkhede: Risk management is evolving very rapidly, which has been exciting and rewarding at the same time, but over the next three to five years I think we should all expect to see a couple of major shifts.

First is hyper personalized real time risk decisions. So traditional risk models will give way to real time AI driven decisions that dynamically adapt to each customer's risk profile. So instead of static rules, just static rules, financial institutions will use live, behavioral financial data to assess risk at every touch point, from onboarding to trade decisions to fraud detection.

The second trend, I think is happening, and will happen even more rapidly, is AI versus AI, the battle against AI powered fraud. So as fraudsters increasingly leverage generative AI to create synthetic identities, deep fake scams, financial institutions will need even more advanced AI to detect and counter these attacks.

In my experience, that is a really smart combination of traditional predictive machine learning models as well as generative AI models, which we found to be really effective methodology in this battle against AI powered fraud.

And last but not least, on the compliance front, unified compliance, embedded risk platforms. So the era of these fragmented risk solutions is coming to an end. We are moving to an era of consolidation. Institutions will move towards unified platforms that seamlessly integrate fraud, AML, credit risk decisioning into a single system, and this will allow better cross functional oversight, automated regulatory compliance, even more efficient risk management without increasing operational costs.

Ultimately, I think risk management is shifting from being a defensive function to a strategic enabler. It's going to help businesses grow faster, safer, more inclusively, while staying ahead of evolving threats. And as consumers, we're going to appreciate better safeguards and much better customer experiences, which is just a win win for everyone.

Nik Milanovic: So how should financial institutions be thinking now if they're looking three to five years down the road, and want to be able to better prepare for emerging threats?

Neha Narkhede: Any financial institution should take a proactive and AI driven approach to risk management to really be able to stay ahead of emerging threats, which is responding to sort of historical threats that they might have seen more commonly in the past.

I think what they should be doing now is adopting real time AI powered decisions. Fraudsters are using AI and automation to scale attacks. So relying on static rules outdated models is just no longer enough. So you need a combination of data sources. You need ML Model Driven risk scoring to detect these evolving threats instantly. You need real time feedback loops to retrain your ML models, which is tough from a technology perspective, we're definitely seeing it's possible, as we've seen at Oscilar.

The second thing that these institutions are going to need to do is break down silos. Many banks still rely on fragmented, disconnected tools. A unified platform is going to give them the oversight and real time collaboration across different risk functions and enhanced AI explainability and regulatory readiness, like we talked about. Regulators are scrutinizing AI decision making, financial institutions must ensure that the risk models are explainable, auditable, compliant with evolving standards, and just be future proof with adaptive infrastructure.

So this risk landscape is just moving too fast for static systems. Institutions will need flexible AI, API driven platforms that can quickly adapt to new threats, integrate emerging data sources, continuously refine these models in real time.

So by taking these steps now and just accepting these trends, financial institutions can stay ahead of emerging threats, reducing fraud, improving compliance efficiency, and I think just building a safer, more resilient financial ecosystem at the end of the day.

Nik Milanovic: Neha, thank you again so much for making the time to talk. It's really great to have you at This Week in Fintech. You know, all these conversations go in different directions. I feel like sometimes we talk about personal life outside of work, sometimes we talk about company history. But I'm glad that we were able to really just dig down into the specifics in this space that I think many, even in financial services, do not understand at a deep level.

Keep reading

Amy Sariego

Feb 12, 2025

Oscilar Founder Neha Narkhede: Eliminating Risk With AI

Patrick Rikli

Jan 30, 2025

Oscilar and Spinwheel Announce Strategic Partnership to Enhance AI-Driven Debt Management and Risk Decisioning

oscilar sentilink flexcar
oscilar sentilink flexcar
oscilar sentilink flexcar

Amy Sariego

Jan 10, 2025

Flexcar Reduces Fraud, Repossessions, and Delinquency Rates with Oscilar and SentiLink

Subscribe to email updates
Subscribe to email updates

Stay in the know and learn about the latest trends in fraud, credit, and compliance risk.

Stay in the know and learn about the latest trends in fraud, credit, and compliance risk.

Meet our Platform

Device and Behavioral Intelligence

Onboarding

Consumer Onboarding

Business Onboarding

Merchant Onboarding

Fraud

Account Opening Fraud

Transaction Fraud

Account Takeover

First-Party Fraud

Scams

Credit

B2C Underwriting

B2B Underwriting

Portfolio Management

Collections

Compliance

AML for Sponsor Banks

AML for Fintechs

Industries

Fintechs

Banks

Credit Unions

Sponsor Banks

Customer Stories

Flexcar

Coast

Transend

Super

Dibsy

Fluz

Resources

RiskCon

Blog

Webinars

Ebooks

Marketplace

Company

About

Contact Us

Press

Careers

Security

© 2025 Oscilar. All rights reserved.

Terms of Service

Privacy Policy

Cookies

Meet our Platform

Device and Behavioral Intelligence

Onboarding

Consumer Onboarding

Business Onboarding

Merchant Onboarding

Fraud

Account Opening Fraud

Transaction Fraud

Account Takeover

First-Party Fraud

Scams

Credit

B2C Underwriting

B2B Underwriting

Portfolio Management

Collections

Compliance

AML for Sponsor Banks

AML for Fintechs

Industries

Fintechs

Banks

Credit Unions

Sponsor Banks

Customer Stories

Flexcar

Coast

Transend

Super

Dibsy

Fluz

Resources

RiskCon

Blog

Webinars

Ebooks

Marketplace

Company

About

Contact Us

Press

Careers

Security

© 2025 Oscilar. All rights reserved.

Terms of Service

Privacy Policy

Cookies

Meet our Platform

Device and Behavioral Intelligence

Onboarding

Consumer Onboarding

Business Onboarding

Merchant Onboarding

Fraud

Account Opening Fraud

Transaction Fraud

Account Takeover

First-Party Fraud

Scams

Credit

B2C Underwriting

B2B Underwriting

Portfolio Management

Collections

Compliance

AML for Sponsor Banks

AML for Fintechs

Industries

Fintechs

Banks

Credit Unions

Sponsor Banks

Customer Stories

Flexcar

Coast

Transend

Super

Dibsy

Fluz

Resources

RiskCon

Blog

Webinars

Ebooks

Marketplace

Company

About

Contact Us

Press

Careers

Security

© 2025 Oscilar. All rights reserved.

Terms of Service

Privacy Policy

Cookies

© 2024 Oscilar. All rights reserved