Why I folded Orionfold

People ask me what Orionfold means, and why I started it. The honest answer runs back twenty-five years. So let me tell it that way, because the whole arc points at one simple idea.

A long run at making powerful tech reachable

In the late 1990s I worked at Xerox PARC, and helped build one of India’s first systems for storing and sharing documents across distance. Even then, the work that pulled at me was the same work I do now. Take something powerful and far away, and put it within reach of more people.

I kept chasing that one thread for years. I helped teach machines to sort information back when that was rare and strange. I led teams building software for large companies. Different jobs, one thread running through all of them.

Scale, and what it taught me

In 2017 I joined Amazon, and I would end up staying for nearly nine years. In the first stretch I helped grow its cloud business in India, from about 5 million dollars to 150 million in three years. Then came something I will never forget. I led the team behind India’s pandemic response, a system that reached 100 million people in 40 days. It is work I am still humbled to have been part of.

That taught me what it truly takes for technology to reach people, fast, at the scale of a whole nation. And it left me with a conviction I could not shake. Power like that should not only live inside big institutions. It should be reachable by anyone.

At the center of big AI

The next years put me at the very center of how big AI gets made. I worked on the conversational AI behind Alexa, where in two quarters I grew it from 1 language to 9, on track to all 16, with a 14-fold jump in traffic. I joined Amazon AGI, the group building Amazon’s own foundation models, on how they learn from examples. Then I led a push to bring generative AI across Amazon’s retail store experience.

I saw how big AI is built, up close. I also saw the cost of it. How much it takes to run. How locked in it is. And, quietly, who gets left outside.

The thing I noticed on the frontier

Most recently I have led the solutions work helping frontier AI companies build on AWS, partnering with labs like Anthropic and chipmakers like NVIDIA. Some of the best-funded, fastest-moving teams in the world.

And I noticed something that would not let me go. The advantage was shrinking. The gap between what those teams could do and what one careful person could do was closing, fast. The stories before this one are my proof. Open models caught up. Frontier thinking got cheap. A real product fit inside a weekend. A model learned my field on a small machine in my own room.

If a single person can now build on the frontier, then the most useful thing I could do was not to help the giants get a little bigger. It was to help the individual. The solo builder. The small, private team. The person standing outside.

So I left

So I left. After nearly nine years at Amazon, I walked away from the center of it all. Not because anything was wrong, but because of a conviction I could no longer set aside. The best use of everything I had learned was to put it in the hands of the individual builder, not to help the giants grow a little larger.

As I write this, on the second of June, 2026, I am starting Orionfold officially. One person. One desk. Out in the open.

What it means to fold

So I folded. I took twenty-five years, and the whole sprawling frontier of AI, and I folded them down into one small studio on one desk.

That is the name. Orion, for the stars, for the frontier above us. Fold, because the whole thing now folds down small enough for one person to hold. The frontier, folded down to your desk.

Orionfold builds open AI software, custom models, and the playbooks to run them. All of it local and private by default, made to run on your own computer, not rented from someone else’s cloud. The point is simple. Grow 10x with AI, on hardware you own, with tools you control.

Why it has to be this way

Everything I have learned says the same thing. Powerful technology matters most when it reaches the person at the edge, not just the institution at the center. I spent a career carrying it toward the center. Orionfold is me carrying it the other way.

If you want the short version of who I am, it lives on the about page.

There is one hard problem standing in the way, though, and it is not the one most people expect. Most teams are pouring money into AI and getting almost nothing back: Why most teams get nothing from AI.