How Capital One Delivers Multi-Agent Systems with Rashmi Shetty

**Sam Charrington** (0:01)
Capital One's tech team isn't just talking about multi-agentic AI, they already deployed one. It's called Chat Concierge and it's simplifying car shopping.
Using self-reflection and layered reasoning with live API checks, it doesn't just help buyers find the car they love, it helps schedule the test drive, get approved for financing, and estimate trade-in value. Advanced, intuitive, and deployed. That's how they stack. That's technology at Capital One. To learn more about AI at Capital One, visit capitalone.com/tech/ai.

**Rashmi Shetty** (0:38)
We moved from a classic ML world to a world where we have LLMs generating responses and now we want to move on to a world where actions need to be taken. And when the problem that we are working on is a complex one, that's where multi-agentic comes into play.

**Sam Charrington** (1:08)
All right, everyone, welcome to another episode of The TWIML AI Podcast. I am your host, Sam Charrington.
Today, I'm joined by Rashmi Shetty. Rashmi is Senior Director of Enterprise Generative AI Platform at Capital One. Rashmi, welcome to the podcast.

**Rashmi Shetty** (1:23)
Thank you, Sam, thanks for having me here.

**Sam Charrington** (1:25)
Thanks so much for joining us. We're gonna be digging into the why and how of multi-agentic AI at Capital One. To get us started, I'd love to have you tell us a little bit about your personal journey.
How did you get to where you are?

**Rashmi Shetty** (1:39)
Yeah, so I think my personal journey is something many folks might relate to.
It's funny how you ask me about my personal journey in a day and age where agentic AI is in its explosive form. So we have stopped pausing to think about it. How did we get here? So, yeah, I mean, yeah, it's been more or less organic and natural evolution in my specific case. But if you ask me how can I pin it to a specific phase of my life, I wouldn't go back to my thesis work.
The topic of my thesis was around pervasive computing. Very interestingly, it was around perceiving your environment, getting signals from your environment, taking context aware decisions in real time, and having actuators trigger actions to change that environment based on your specific goals. It was a physical world in which this was manifesting. But in a sense, if I think about it, the principles were very much the same. It's about distributed intelligence. It's about intelligence being embedded in the system that you're living and operating in. And if you think about the governing principles, where we were working on the design at that point in time, which my thesis topic was around, was specifically around decisioning systems, real-time intelligence. And that's where we are today. I think moving from academia to the industry just kind of taught me how to scale that in an enterprise, in an enterprise world. So the next phase was explosion of data, building out AIML pipelines, the AutoML platform journey, where you need to operationalize pipelines at scale. And here we are now in the agentic world, where we are moving beyond that decisioning systems and actually taking actions, which feels like a closed closure of the loop for me.

**Sam Charrington** (3:52)
How long has Capital One had a generative AI title or role?

**Rashmi Shetty** (3:59)
So generatively, we have been, for those who don't know much about Capital One, we've always been at the forefront of modernization.
So we've been in the AIML journey for the past decade, even governed data, even before that, we've been one of the first banks to go on modern cloud platforms. Similar to that, the GenE AI journey began a few years ago, like early 2023, was when we embarked on this mission. We had our first pilot, agent assist pilot, which went live in 2023 And then now we've had a GenE AI presence and now an agentic presence in the past two years.

**Sam Charrington** (4:43)
Talk a little bit about kind of the motivation for multi-agentic systems. Where does the need for multi come in?

**Rashmi Shetty** (4:56)
So here is the philosophy behind it, right? We moved from a classic ML world to a world where we have LLMs generating responses. And now we want to move on to a world where actions need to be taken, specific goal-oriented actions need to be taken. And when the problem that we are working on is a complex one with multifaceted aspects associated with it, that's where multi-agentic comes into place. So basically, we have a large complex goal, which we have to break down into specific steps.
And each step is basically narrowed to a specific agent. And that agent is tasked with the goal of executing that specific task. And then you move on to the next agent. So a multi-agentic system or architect or orchestrated system comes into play when you have a large complex goal, which can be broken into steps. And you can take this entire system to fruition of that one complex goal that you have. So that has been the founding principle behind our backing of multi-agentic architectures.