AWS Financial Services Symposium 2025 | Peyman Parsi, MongoDB

Clips
News
More from AWS Financial Services Symposium 2025

Peyman Parsi

Senior Principal, Financial Services Industry Solutions, Americas

MongoDB

Inside the AWS Financial Services’ AI playbook: 8 expert conversations from labs to ledgers

AI for financial services is no longer a distant vision — it’s reshaping how the industry thinks about risk, regulation, customer experience and innovation itself. From streamlining fraud detection to modernizing data architectures, financial institutions are leveraging AI not just to keep up, but to lead.As enterprise-grade tools mature and cloud platforms such as Amazon Web Services Inc. scale up their offerings, these organizations are deploying AI in tangible, transformative ways — and watching new value vectors emerge in real time.That momentum was on full display during the AWS Financial Services Symposium, where theCUBE captured a series of candid, expert-driven conversations

play_circle_outline MongoDB's modern database for financial services with AI integration

play_circle_outline Innovative Approach to Trust Building and POCs Through Modernization and Generative AI for Data Migration

play_circle_outline Embedding AI everywhere in financial services for vertical use cases

play_circle_outline Maximizing Efficiency: Role of Domain Flexibility and Vertical Targeting in Digital Workers

Info
Transcript

Peyman Parsi, MongoDB

Peyman Parsi

Senior Principal, Financial Services Industry Solutions, Americas MongoDB

At the AWS Financial Services Symposium, Peyman Parsi, senior principal of financial services industry solutions – Americas, at MongoDB, joins Dave Vellante and Scott Hebner of theCUBE Research for an exploration of how AI is redefining the financial services landscape. The discussion highlights how MongoDB’s flexible, data-centric architecture provides the agility financial institutions need to modernize legacy systems and power AI-driven solutions at scale.

Parsi explains how MongoDB is helping financial firms move AI from proof of concept to product... Read more

explore Keep Exploring

What type of database is being provided in the landscape of financial services, and what are some of its key features? add

What is the program called that is being used to modernize and build trust with large banks or large accounts through a pilot process involving qualified applications and generative AI? add

What is the current state of financial services in terms of legacy modernization and how can AI potentially help in this transformation process? add

What is the new opportunity for software companies in targeting labor spend with digital workers instead of IT spend? add

bolt Powered by CUBE AI

Peyman Parsi, MongoDB

search

Dave Vellante

>> Hi, everybody. Welcome back to New York City. This is Dave Vellante with Scott Hebner, my co-host, and you're watching theCUBE's coverage of AWS's Financial Services Symposium. This is the 10th annual. Started out as a small little get-together with a bunch of FS Pros, probably 30 or 40 of them at the Roosevelt Hotel, and it's now exploded into a 6, 700 person event here in New York City in Tribeca. Peyman Parsi is here as a senior principal financial services industry solutions for the Americas at MongoDB. Good to see you, friends of Mongo are friends of theCUBE. Welcome.

Peyman Parsi

>> Yeah, great to be here. Thanks for the opportunity to speak with you.

Dave Vellante

>> Yeah, so give us the update on Mongo generally, but specifically what's happening in financial services and how AI is really proven to be a catalyst for innovation.

Peyman Parsi

>> Yeah, we are providing a modern type of database in the landscape of financial services. We have a broad scope of use cases that we apply to bringing really a new dimension to the data and also integrate that really well with the application practices. So our platform really blurs the line between software development and data. Bring these two pieces together really to expedite and create agility, but at the same time provide a data platform that is flexible and also have ability to adjust the data structure on the fly without any need for shutting down the database. So at the core of it's a document model. We call it document model, but it's not really about PDF documents. It's a JSON format and really integrates well with the entire ecosystem of financial services. As you know, there's a lot of legacy modernization and a lot of changes in financial services, and we will be the optimal choice for all those modernization because we support also data distribution across our platform. At the same time, we support horizontal scalability. Today, the only thing that is going to happen to data is going to grow, and 80% of that data is unstructured data. We are really well positioned to support all those transformations.

Dave Vellante

>> In reality, it's probably 99%, but we've been using that 80% figure for a long time. I've been covering, and theCUBE's have been covering MongoDB since it was called 10gen.

Peyman Parsi

>> Yes.

Dave Vellante

>> We go back a long way. And at the time it was very obvious to us the developer, Affinity. Developers loved that sort of document database and the ease of which they could actually get stuff done and build applications. Specifically the knock on it was, "Oh, it'll never scale." Well, here we are in 2025, you've got bank grade loads using the platform, it scales. I wonder if you could talk about that and maybe share any stories or proof points that best really illustrate your ability to handle the sort of extreme workloads in financial services.

Peyman Parsi

>> Yeah, absolutely. We are in financial services, 9 out of 10 top banks in North America using our product 7 out of top 10 bank globally using our product. So in terms of expansion and number of basically overall exposure globally. So Atlas, which is our cloud platform database is downloaded 195,000 times registration on Atlas and downloads on our community version. So it just shows the scale. And also 70% of Fortune 100 companies running our product. So we really at the core of running mission-critical banking workload across North America and globally.

Scott Hebner

>> These firms are sitting on a ton of POCs, right?

Peyman Parsi

>> Yeah.

Scott Hebner

>> They get to a certain point and then you start getting into issues of trust, compliance, governance, the quality of the data. These firms have to figure out how to get the POC over the line and get it into production and start delivering value.

Peyman Parsi

>> Yeah.

Scott Hebner

>> What are your thoughts on that

Peyman Parsi

>> So we have a very specific program in response to that, it's called Modernization Factory, and we work with the large banks or large accounts to make sure that trust is built through a pilot process that we have in there. So we pick a number of different qualified application and try to help transforming them to the new data model. And also at the same time we use generative AI in the same process to expedite that process. It's very innovative approach to modernization. We also offer relational migrators part of our offering in terms of that migration. But I think with the AI is a little bit different because AI trust means different thing and hallucination and especially generative AI. With that dimension, we are bringing new capabilities. We just recently acquired Voyage AI and we are bringing top ranking, embedding and re-ranking and domain specific models to our offering because we believe these kind of capabilities belong to the data layer and we already have most loved vector search as part of our offering. We are bringing embedding and re-ranking also to that pack to increase the trust, enhance the trust, and also reduce the hallucination.

Dave Vellante

>> The amazing thing about your vector searches, it was the most loved vector search on multiple surveys before it even went GA.

Peyman Parsi

>> Yeah.

Dave Vellante

>> Remember that? It was two years ago. I remember, we were at Reinvent and I remember reading, I forget which outfit put it out there, but it was one of the developer platforms.

Peyman Parsi

>> Yeah, Ritual. Yeah.

Dave Vellante

>> Ritual. It was you were up to the right, you had Milvus, you had Pinecone, but you were by far the...

Peyman Parsi

>> Highest NPS....

Dave Vellante

>> the highest NPS. I was like, "Whoa, they haven't even announced it yet." I mean it wasn't even really publicly available. So what's been the uptake since then? And it sounds like you've advanced the platform?

Peyman Parsi

>> Yeah, in financial services. There is a lot of discussions going on in past couple of years I would say around our vector search capability because there's a lot of patterns and a lot of behavior in financial services that vector search can help to determine and based on those patterns like fraud detection or customer behavior, you can do more, I would say higher quality fraud detection. It doesn't mean that the ML version of fraud detection or rule base of fraud detection was there has to be throw away. This is another layer of quality on top of that which brings false positive down and also increase the quality. So our customers realize that these sort of capabilities are really changing the game for them and it was not possible before. So that's just one example of it. But there are many example of how that vector search apply in financial search.

Scott Hebner

>> So you're using the vector search in the context of RAG?

Peyman Parsi

>> That's right.

Scott Hebner

>> And then applying the rank, the ranking algorithms to get the most Relevant.

Peyman Parsi

>> That's correct.

Scott Hebner

>> Not right. And is there semantics in the middle of this? And how does this all come together?

Peyman Parsi

>> It's a semantic search at the core of it because majority of unstructured data where we basically embed them, putting it in the database, when we run a vector search on top of that, you get much more relevant result. By re-ranking them, not only we have a better quality data, but we also increase the trust because those more relevant search results will come on top. But that's not at the end of the game because sometime in specific cases in financial services, for example, if you're dealing with financial news or financial instrument, there are models that are specific to that domain. So that's what we do with Voyage AI. We are bringing those domain specific models like in financial services, in law for code generation that just increase the quality even one more level to build that.

Scott Hebner

>> So you take all of that capability and then you have the guard rails.

Peyman Parsi

>> That's right.

Scott Hebner

>> This is how you're helping lower the incidents of hallucination bias, things of that nature-

Peyman Parsi

>> That's-...

Scott Hebner

>> obviously it's critical in this industry.

Peyman Parsi

>> Yeah, and Agentic right now there's a heightened level of trust that needs to be made because if you're talking about decision-making and critical decision-making in financial services, if the agent does something wrong, the impact could be catastrophic. So even the importance of that trust is higher.

Dave Vellante

>> So we tend to lump as an industry as pundits, whether it's AI, machine learning, GenAI, Agentic, all kind of automation. But when I think of some of the use cases for Agentic specifically, where's the fit? I mean, think about popular FS use cases, treasury ops, algorithmic trading as an example, or fraud detection. I mean fraud detection, we were doing pretty well with fraud. Maybe it's triage for fraud. Where does Agentic fit relative to the traditional ML, AI?

Peyman Parsi

>> I think in the banking or financial industry, because there's a lot of processes and there's a lot of gain to be made out of process efficiencies. So that process automation with some sort of decision-making integrated with RAG and relevant data has a lot of potentials. That's one of the areas a lot of our customers are looking at. But beyond that, we just launched even last week a demo in wealth management where there are three agents trying to help a portfolio manager managing a investment portfolio by, it's a human-in-the-loop solution, though still industry is not ready for financial decision to be fully automated. We are getting there. So a lot of enhancement in FinTech and startups, but a banking industry. So we still need that aspect of human-in-the-loop. We just launched a demo to show capabilities around market analysis and also macro level market information. There's a lot of information in earnings report or unstructured data or social media that these agents can bring in. In the form of lower risks because they provide insight to the portfolio manager at this point. And we are hoping at some point this is going to be more automation and more autonomy.

Dave Vellante

>> Interesting hearing you talk about the wealth management application. We can all relate to that, right? I mean, I could write the programs today, the historical ones they use and they do the Monte Carlo modeling and how much money they have? How long are you going to live? What's the market going to do? Let's run a zillion scenarios and tell you when you run out of money. And it's like, "Eh." So is what you're describing a significant scale up from that?

Peyman Parsi

>> Oh yeah, definitely. And I think when you look at technology behind that, part of that is a vector search and embedding and re-ranking all those technologies that we point out. But the other part is ability to make decision and plan by these agents and the trust that needs to be there in order to do that. So hallucination is one aspect of it, but there are new concept in Agentic coming to the play currently experimental, but the cause and effect, for example, or other sophisticated decision-making character of these agents would really change the game in that-

Dave Vellante

>> And real-time what ifs would be kind of interesting to be able to have a conversation with the AI to say, okay, what if I took on more risk? What's the best case? Worst case, now you're getting best case, worst case, but you don't have the granularity to really dramatically test different scenarios. You run out of gas and run out of meeting time before and you say, we'll worry about that next year. We'll follow up on that. But so if I can have an ongoing interaction with AI around that, and causal, one of your favorite topics, you came and just mentioned plays into that.

Scott Hebner

>> I love that you brought up cause and effect. Yeah, I've been very passionate about that being the next step because I think what you're describing here in a more, let's say holistic is we've been using, generative AI to create content, analyze information and to make some predictions. And what's really shifting here is we're going from predictions to judgments.

Peyman Parsi

>> That's right.

Scott Hebner

>> Judgments is a whole new game because it has to help you understand consequences. You're not going to understand consequences unless you understand cause and effect. That's correct. Which gets you to semantical and causal reasoning and causal AI and all that. So it sounds like that's where this is all heading with you guys.

Peyman Parsi

>> Yeah. Absolutely.

Scott Hebner

>> And we've been saying is if you can solve it in the financial services industry, it's going to apply in a lot of other industries. Because this is really hard stuff here with a lot of regulations and compliance and policy.

Peyman Parsi

>> So we want to be ultimate basically choice of data for AI, Workload and making developers' experience in that journey less cumbersome, more know friction less and also easier to integrate. But at the same time, the core foundation of our platform provides that horizontal scalability and performance and real-time availability of data at a better cost performance, basically combination to be able to enable AI workload with agility and with the quality that required. We also invest a lot on cyber security, making sure our solution is robust and bringing innovation to cyber security aspect of our product as well.

Scott Hebner

>> You mentioned, Dave, the importance of interrogating the model. What if this, do counterfactual, all that? The other aspect, other side of the coin of that is explainability, the more it can explain to you why this is the right decision to make given other alternatives. And then explainability is the other angle that it sounds like when you pull all the things that you have talked about so far together, it's going to give you better decision intelligence, it's going to give you more explainability.

Peyman Parsi

>> Yeah, the explainability is really important in financial services because there's a lot of, I would say hand-holding between the processes. Sometimes that flow of what happened to the data get lost and being able to find where the data was originated, what's the lineage of that data and what is the business meaning? Business' story behind the data is really the explanation would help. And we also see a lot of cases around summarization and bringing more intelligence to different workflows. For example, fraud management used to be very binary, used to be yes or no. Now you have explanation of why it's flagged as a fraud. And also sometimes it's again, human-in-the-loop, get that information and take it to the next level.

Dave Vellante

>> John Furrier calls AI everything everywhere, all at once. I collect the movie. And so if you think about going out to the end of the decade, let's say, will AI be embedded everywhere, all at once within financial services and what's the business case for that and how will it get there?

Peyman Parsi

>> I think it will get there, but financial services right now in a very, I would say continuous cycle of transformation in terms of legacy modernization. So that's probably competing with the AI being used if there are solutions that can help modernization with AI, as I mentioned, our ModFactory is one of the offering of that. But also not only would help modernize better and faster, but also bring AI capability to be realized much, much faster. Right now what we see AI capabilities across financial services is mostly around corporate GPTs or knowledge base, supports, whether it's offering-

Dave Vellante

>> Kind of rag-based chatbot.

Peyman Parsi

>> Rag-based chatbot, but I think the vertical use cases when you see the real hockey stick of the impact and those vertical use cases all mostly under proof of concept right now. So we'll see in next couple of years, they come to the play and combining with Agentic there's huge potential.

Dave Vellante

>> Well, I'm sorry to interrupt, but right now it's picks and shovels, it's Nvidia, it's Broadcom, it's the cloud players. I've heard Dave lay out his vision and it's that application layer is where the real enterprise value is going to be realized.

Peyman Parsi

>> Yeah and that's part of our core value prop. We always have developer's life cycle in mind and we wanted to make sure that journey is easier and more effective and agile. And also we provide the services that our customer want to be part of the data layer. We really listen to our customers. If you look at the history of Mongo, we brought in capabilities that our customers are asking and it really fit the data layer because it makes the workflow much easier and more integrated and seamless.

Scott Hebner

>> Your point really resonated with me about domain specific vertical being the opportunity. Because a lot of people out there, I think I'm becoming one of them, that is believing that the new opportunity for these software companies is targeting the labor spend with digital workers, not the IT spend. And when you think about how you help labor in workers with digital agents, you get very domain specific very quickly, how do I manage contracts in the financial services industry that have to deal with European regulations, right?

Peyman Parsi

>> Yeah. That's the next level for sure.

Scott Hebner

>> The SaaS apps maybe try to go after that, but for agents you can go directly at that level of added value and you're topping into labor spend. It's a whole new market.

Peyman Parsi

>> Exactly. And that's part of why we acquired Voyage AI to have that flexibility as part of our platform. We're also investing a lot in integrating that within the platform, but also providing auto embedding capability for text that is integrated with vector search. So that way is come very neat because you don't need to go elsewhere to find the models. And we also keep it open, if you want to bring in your models or you want to interact with other models, it is seamless and easy, but have ability to apply those models that are industry specific on the fly within the database realm. It's really game changer for us.

Dave Vellante

>> Well, Peyman, thank you so much. This has been fascinating. Really appreciate you coming on theCUBE.

Peyman Parsi

>> Absolutely. Great talking to you. Thank you.

Dave Vellante

>> You're welcome. All right. This is Dave Vellante for Scott Hebner and you're watching theCUBE's coverage of AWS Financial Services Symposium. We're here in New York City in Tribeca. We'll be right back right after this short break.