Frank Petterson, Uniphore CTPO on Enterprise AI and Product Leadership

**Grace Gong** (0:02)
Hi, Frank. Welcome to Venture with Grace.

**Frank Petterson** (0:05)
Thank you. Thank you for having me.

**Grace Gong** (0:08)
I'm excited to be here. Me too. I'm so excited about our conversation. I'm so sorry. I don't know why there's a little echo on our stream. So I asked Frank to mute when we're not talking, but thank you so much, Frank, for coming. I'm so thrilled about our conversation. Before we start off the show, I want to give a quick shout out to our amazing sponsor. This episode is brought to you by Nebius, the ultimate cloud for AI innovators. Nebius provides AI infrastructure you can count on, combining reliability and speed with flexibility and engineering support unmatched by hyperscalers. AI leaders like Meta, Shopify, and Higgsphere already partnered with Nebius to run their AI workloads. Plus, venture-backed startup can save up to 150K on compute costs when they apply for access. Visit nebius.com or nebius.com/startup to learn more.
I want to start from your background. You went to school at Stanford and you have worked at Meta before joining Uniform.
Why don't we start with what were some core lessons that you've learned early on in your career that shape you into who you are today?

**Frank Petterson** (1:12)
Yeah. Like you said, I came to the Bay Area for Stanford. I did a Ph.D. there.
The main thing that I was actually focused on even back then was computation of large systems. We were doing a lot of work with GPUs at the time. This was when the first programmable GPUs came out. And that was an exciting opportunity, right? Like GPUs really fundamentally changed what you were able to do.
But then my career kind of left me all over the place. Like you said, I went to, well, after Stanford, I went to a company called Industrial Light & Magic. And then I went to Google, YouTube, AliveCor, Meta, and then now Uniphore. And yeah, it's been kind of a journey across a few different fields.

**Grace Gong** (2:06)
For sure.
I think like you, maybe we could start from like, you know, Uniphore as a company. Obviously, your CEO was on our podcast previously. But for those who don't know, maybe we could start from the background of Uniphore. Obviously, you guys have grown, like, you know, you guys grow over 100% since we last tried. Maybe we could start from, you know, what is Uniphore and what is like business AI company, like, in general, like, you know, the business of it, of Uniphore today.

**Frank Petterson** (2:41)
Yeah, so Uniphore definitely has changed over time. It's, I believe, you know, getting up on kind of, you know, almost a teenager now, I guess it is a teenager now. So it's not a new one. But the technology that Uniphore has been developing in the last kind of handful of years is definitely moving towards the AI or with the AI wave. It's always been a conversational AI company. But now we're looking more at what is essentially an end to end data and AI orchestration platform. So we actually help companies gentify their business. That means that they can use their data, their business context to fine tune models. We are heavily invested in SLMs as the main method. Then build agents. We serve about 2,500 companies. Many of them are Fortune 100 companies. We primarily focus on large enterprise.
The fundamental question for Uniphore has always been, or has recently been, why hasn't every enterprise gone autonomous and agentic? Obviously, AI is really good. We've all seen it do amazing things. We use it in our private lives.
We know that it's not necessarily about the model. The models are really good. They're going to get even better. In some sense, models are somewhat of a commodity at this point.
If you look at the frontier models, they're maybe six months ahead of the open models. We've all seen like Kimi, GLM, DeepSeek, and all of those come out. They're competitive now. The models are everywhere. The models are great. Why don't businesses use AI more? Why do we see things like only a handful of percent of AI deployments are actually successful within the enterprise?
We fundamentally believe that that's due to the lack of good orchestration layers. You all ultimately need to get your data, your context, your models, your agents all working together, and that's what we provide. We think that the value is going to live at that layer, and that we can really help businesses with that. For sure.

**Grace Gong** (5:20)
I agree with you on the small language model versus large language model part. Now, you for choose the industry-specific small language model, and I guess what would you say would be the trade-offs for most, whether it's accuracy, cost, latency, privacy, and control. Obviously, I feel like each company should control their own data since I don't think everyone wanted to hand off their own data into some of the large models.