Ignite Startups: How VoiceBit Is Transforming Restaurant Ordering with AI with Jay Patel | Ep303

Jay Patel (0:00)

One thing that a lot of people, a lot of restaurant owners might not know, is that 70% of their business is being run by their core loyal customers. That is like 30%.

You have to protect them. And there is no reason for them to go to this third party or anywhere else and lose margin on it. You need to own them, you need to provide loyalty benefits, you need to provide personalized assistance, all those things. And so those are the customers who are probably running your business, and you have to protect them.

Brian Bell (0:52)

Hey, everyone. Welcome back to the Ignite Podcast. Today, I am delighted to have Jay Patel on the mic. He's the co-founder and CTO of VoiceBit, and fair disclosure, a portfolio company. It's an AI ordering platform that helps restaurants win back the phone, web, and delivery orders that's been leading them to third-party apps. Before he started VoiceBit, he spent about a year leading Azure support AI work at Microsoft, where his team used generative AI to deflect thousands of support cases. So there's definitely some transfer learning there to use a machine learning term. VoiceBit came out of Berkeley SkyDeck Go Bears and just closed an oversubscribed pre-seed round led by 1984 VC. Thanks for coming on, Jay.

Jay Patel (1:28)

Thanks for having me.

Brian Bell (1:28)

Good to see you again. Well, I usually kick it off with the question, what is your origin story? What's your background?

Jay Patel (1:35)

My name is Jay. I'm originally from India. I self-taught myself programming when I was in school.

It was pretty interesting because back in those days, computers were used to call us a brain.

It was always intriguing to me what exactly goes on inside those boxes.

So learning how to code actually gave me that appreciation on how these things work. And because of that, I just did my bachelor's in computer science. I learned a lot of things about this. Sorry, I learned a lot about all of this theory of computations, the operating systems, the information theory, all those advanced topics. So after that, I worked probably three years as a software engineer back in India. And that's when I learned a lot about practical software engineering. Like how the things work in production environment, how the things is built, maintained, deployed in production environment. And that's when I learned, that's when I actually realized how it actually separates from the real world and what I learned back in the theory.

After that, I came to the United States to do my masters in computer science. And I think one of the most interesting subject that I learned here in Cal State LA was, I think two of them. One was the information theory and another one was the reinforcement learning. And I'm talking about 2019 because it was way before all of this AI boom started. And the moment I learned about this Markov decision process, MDP, I'm not sure if you know about this, but it's, it's the center point about MDP is that it's just whatever state you are in right now. That's what matters. Nothing in the past matters. And so the current state holds the entire history. And that's all you have to know about in order to predict the next state, the next base state. And so that was the, that was the aha moment for me. It's like, I think it totally makes sense. Why did I not even think about it before? But after that, I joined Visa to do my, again, resume my professional computer science or the software engineering career. There I was working on the loyalty and benefits platform. So you see all these offers in your cheers. I'm not sure if you use cheers, but you see the offers on your cheers app. Those are the ones being targeted to you by the loyalty platform. And so the way it works is that it analyzes your spending habits and then segments you. And then that's how they target you. And so it was amazing experience there, like how to process petabytes of data and trying to figure out how the things work in terms of distributed computing. After that, I moved to Microsoft to work on Azure. And it was a completely different landscape there. Like the services that we were maintaining were like a global scale, like billions of requests every day. And the platform that I was working on mainly was dealing with the Azure Support AI. And what that means is, let's say if you have any issues with your subscription on the Azure, let's say you are not able to see the data in your storage account. Usually, those require some specific set of steps. They are mostly pretty fine. Like you don't really need to be a rocket scientist to know that. However, when you are doing it for the first time, you definitely need to know about those things. But when you think about the Azure scale, it's like millions of requests like this. And you just cannot afford to have an engineer working on those requests. And so the idea was that you need to figure out based on the request from the user, if we already know such solutions from our knowledge base, then how can we even give those solutions to the user rather than them creating a ticket and then having an engineer work on it and then back and forth. So that was the thing that I worked there. But yeah, and after that, I started the VoiceBit.