Thesis is building an AI lab for all. We give you a unified environment for experiments, models, datasets, and agents that accelerate your AI research and development (R&D).
We are democratizing intelligence building. Why should it be the case that the future direction of one of the most important technologies of our time is spearheaded by only a few frontier labs? It's our fundamental conviction that AI R&D should be widespread: permeating every facet of scientific progress, for the good of humankind. Neo labs have shown us the promise of AI applied to every domain imaginable: from helping discover new life-saving drugs to creating the brain behind the robots that will power the physical economy.
But an AI lab's rate of progress is fundamentally limited by the breakthroughs they make everyday. It is left to chance. At Thesis, we are ushering in a new era of discovery, one facilitated by our human-in-the-loop lab. It helps you surface insights about your experiments and draw connections to existing work that you might have otherwise missed. It monitors your machine learning (ML) experiments for you, so if something pathological happens at 3am, you don't waste millions in compute. We are giving everyone access to an AI lab because we believe the most important problems that AI can be applied to are yet to be solved.
We view human AI researchers and engineers as a crucial, symbiotic part of the lab. They guide the big picture ideas and the research direction. But they no longer have to be bogged down by random out-of-memory errors because batch size was misconfigured. We offer a system that track ML experiments for you, proposes promosing directions, and self-heals as errors arise.
In November of 2025, we achieved state-of-the-art on Open AI's Machine Learning Engineering benchmark, the first real test of whether AI agents can train machine learning models. Since then, we have been excited to give these capabilities to the rest of the world and see what scifi-level breakthroughs are made!
Our Ask
We're looking for:
- AI researchers and engineers who want to move faster
- Teams drowning in experiments, runs, and half-remembered insights
- Early users willing to give blunt feedback
Stay tuned for more updates as we continue to build and grow.