Young builder’s notes: Sharpening the axe

We were introduced on the Bay Bridge on my way to San Francisco. Windows down, enjoying the salty air, I looked to my left and read a billboard for Nebius — “AI runs better on us.” As a twenty-year-old college student in tech, I was immediately hooked.

Intrigue turned into research, where I learned about the recently launched Token Factory. Research turned into conviction, where I became a proponent of the “full-stack” approach to inference. Conviction turned into action after I watched an interview with the Chief Business Officer of Nebius and was drawn to Roman Chernin’s ambition and “go get it” attitude.

For the past two years, I’ve been developing an AI marketing startup, teaching myself how to build large inference systems with nothing but ChatGPT. This also taught me that we live in a time when a kid with a laptop can do just about anything. With that in mind, I didn’t want to watch Nebius just from the sidelines.

Somehow, I knew: if anyone would respond to a college student’s letter, it would be Roman. I found his email and made my case, signing off with: “Nebius is building the future of AI infrastructure, and I want to be part of it.” I didn’t expect a reply, but that wasn’t a reason not to shoot my shot. Less than twelve hours later: “We’d love to convert your energy into value for business.”

There is an undeniable pattern in this industry. Woosuk Kwon, a PhD student at UC Berkeley, built vLLM — the inference engine that quietly serves most of the world’s AI. Michael Truell, twenty-five years old, built Cursor, an AI coding platform that is reshaping how software gets written. Technologies so disruptive they rewired entire industries, built by people my age, often before anyone asked them to.

Of course, the foundation of the AI we know today wasn’t created by them. ChatGPTs and Claudes are the product of decades of quiet, compounding work. Many of the AI pioneers who made it possible are no longer here to see the impact. But while our generation did not invent the technology, we bear responsibility for its direction. And the weight of that responsibility is arriving faster than any of us expected.

Now, as part of the Nebius Token Factory team, I’m building industry-leading AI inference systems on the world’s most advanced hardware. There is not a day when I am not on the cutting edge — and not a day when I take that for granted.

I am a solutions architect, and my main work consists of running proofs of concept for potential customers. We meet, establish needs and requirements, and then I head to the forge — spending my time optimizing, customizing, and deploying models to fit their exact needs. This is where the battle is won: each client has a unique use case with specific challenges, so each cycle has its own set of problems to solve.

Here, in the Token Factory, new skills, tools, and rules arrive daily, changing the playing field. Sitting hyper-focused at my desk, there are often days devoted entirely to learning. The most complex challenges I face are not about execution, but about sharpening the axe.

In an industry moving this fast, it’s easy to feel humbled by the pace. I often do, but I think that’s the point. The people who built the foundation of AI didn’t know where it was going either. They just kept building. I’m doing the same, while only beginning to realize how quickly the ground can shift.