Ignite AI: The Future of Voice AI Testing and Self-Improving Agents with Tarush Agarwal | Ep284

**Brian Bell** (0:00)
What's the best advice you ever got that you initially ignored?

**Tarush Agarwal** (0:03)
I think focusing on your problem. We could have very well gone and built a voice agent platform. We could have gone and built models. We could have done simulations for everything. We could have done built a generic tracing platform. But what we really wanted to solve for is conversational agents.
And that's the best piece of advice. We just kept our focus on working really closely with our customers, knowing them, knowing their problems deeply, knowing how they use the platform and building from there, not coming up top down and be like, oh, hey, let's go build this, right? That doesn't work. And coming from a job, right? But usually, you're told what to do versus having to figure out from ground up, what is working, what is not working, what really even needs to be built. And that piece of advice was definitely one of the key things I hold near India to do.

**Brian Bell** (1:21)
Hey, everyone, welcome back to the Ignite Podcast. Today, we're delighted to have Tarush Agarwal on the program. He's the co-founder and CEO of Cekura, which is building the infrastructure for self-improving voice agents, the verification layer for the world, where agents talking to your customers are increasingly written and rewritten by AI. Thanks for coming on, Tarush.

**Tarush Agarwal** (1:40)
Thanks for having me, Brian. Really appreciate it.

**Brian Bell** (1:42)
Yeah, I'd love to start with your background. What's your origin story?

**Tarush Agarwal** (1:44)
Definitely. So yeah, I grew up in India, as you can tell. I spent most of my adult career there. So I went to college to IIT Bombay. I did major in computer science. So if you know about India or IITs, it's really one of the premier institutes in India. Following my graduation, I very quickly moved on to work at a quant firm in London. I spent a couple of years in London. I started working on low latency quant strategies. So I ended up as a quant developer there, ended up working in a team which works on really low latency quant strategies.
It was quite an exciting time. Definitely, I think I learned a lot. It was my first proper job out of college. I spent time with engineers, really great engineers. Half my team was from Citadel, OptiWare, and half the team went to those places after. So it was a really great team, some of the smartest people I've worked with. I got presented with an opportunity two years into it to move to Chicago. And that was really exciting. It was a very small team of us who were starting the Chicago office. And that came as a very big excitement for us. Mark that. Yeah, it was a big opportunity. I was quite excited because it was a really small team. And that means naturally a lot more ownership, a lot more learning. And yeah, as you can guess, I took the first opportunity to go there. I spent a year there.
It was some of my favorite time spent, right, building and writing code from scratch. What I also realized over there is, okay, like I really enjoy ownership. I like building things from zero to one. And hence I decided to take this plunge. I decided to quit my job and started applying, started hacking. And we got into Y Combinator with one such idea, which is Cekura. And yeah, the two years after that has been a journey, ups and downs.

**Brian Bell** (3:34)
I bet being a founder, is it harder or easier than being a quant dev?

**Tarush Agarwal** (3:39)
I think there's pros and cons. I wouldn't lie.

**Brian Bell** (3:41)
Because the kind of stuff you're building is super technical, right? To build quant systems at a hedge fund is just... Yeah, it's got to be some of the most technical stuff you can do.

**Tarush Agarwal** (3:49)
We were not even thinking milliseconds, we were thinking nanoseconds. Yeah. Really, that was the margin. We were looking at 0.1 bips of movements in the prices. We were looking at 9 to 7 nanoseconds of latency. Wow.

**Brian Bell** (4:04)
Everything is down in C, you're down in almost like the compiler layer trying to get this more efficient.

**Tarush Agarwal** (4:10)
It's definitely a lot of that. It's also a lot of architecture. Do you use radio waves or do you use microwaves? There's also laser coming up.
A lot of architecture, a lot of system design, and of course, the low latency. On top of that, how do you design your algorithms? It's really the full thing.
You have to really understand that it flows end-to-end. You can't just be sitting and touching one piece of code. That's what I really loved about my job as well. I would touch like one strategies. I would go and look into order books, understand how the orders are flowing in, where we're doing well, we're doing poorly. Go dig deeper into that data, find that place where we can improve that one nanosecond. That's really what excites me about my previous job. Yeah, I think I still miss parts of it, but I love what I do currently, so I can't complain.