The Story of Titan ML: Building the Invisible Layer of AI Infrastructure

Sometimes the most important companies emerge not from chasing trends, but from seeing around corners. In a recent episode of Category Visionaries, Meryem Arik shared how Titan ML evolved from an early bet on deep learning’s future to becoming a crucial player in AI infrastructure.

From Physics to Finance to Founding

Meryem’s path to founding Titan ML wasn’t linear. “My undergraduate was in theoretical physics and philosophy. And when I was 21, 22, I really didn’t know what I wanted to do with my life,” she shares. This led her to investment banking at Barclays, where she worked as a rates derivative structurer.

But the entrepreneurial spark was always there. “My father is a Founder, so I’ve always been in an entrepreneurial setting,” Meryem explains. “All of our conversations at the dinner table with both mom and my dad, who worked in the tech industry, were about companies and building things.”

Unlike many founders who rush to start companies immediately after graduation, Meryem took a more measured approach. “I didn’t want to start something without feeling like I had a really good idea and I had validated something that I think could be really meaningful,” she notes, adding, “I think part of the reason why I had that requirement on myself… is because I saw from what my parents went through how difficult it is to build your own business.”

Betting on Deep Learning’s Future

When Titan ML started, they made a crucial bet: “We had a very strong intuition when we started it a couple of years ago that deep learning would be huge,” Meryem recalls. But the market wasn’t convinced. “We had investors telling us that they didn’t think that NLP and language AI would be a big enough market.”

Then ChatGPT happened, and everything changed. “I think what we found in the beginning half of 2023 is that businesses wanted to, quote unquote, ‘do AI,’ but hadn’t figured out how to do it. So there was just like a bit of a, almost a panic.”

The Core Problem

Titan ML identified a fundamental issue in the AI landscape: infrastructure was becoming a bottleneck. “ML engineers are spending way too much of their time on building infrastructure rather than solving the problems that are really core to their business,” Meryem explains. “If we think that AI is going to be as widely adopted as we think it is, then we have a lot of work to do, and we can’t have ML engineers at every single business and every single enterprise building the same infrastructure over and over again.”

Building for the Future

The vision for Titan ML extends far beyond current market dynamics. “We would like Titan ML infrastructure and Titan ML technology to be a core and probably invisible part of almost every single LLM deployment and generative AI deployment,” Meryem shares. “We want it to be analogous to deploying a database when you don’t even really think about it, and it’s just kind of part of your stack.”

Looking ahead to 2024, the company is expanding beyond language models. As Meryem explains, they’re “expanding the takeoff inference server beyond just llms and towards more generative AI modalities. So looking at things like audio and audio models and vision models are things that we’re really looking forward to, that our clients have been asking us for.”

The story of Titan ML is still being written, but its trajectory highlights a crucial truth about building infrastructure companies: sometimes the biggest opportunities come not from following trends, but from solving the fundamental problems that emerge as those trends mature. As AI continues its march toward ubiquity, the companies building the invisible layers that make it all work may end up being the most crucial players of all.