Who is Sergio Charles?

Sergio Charles is the co-founder and CEO of Thesis, an applied AI lab accelerating the frontier of AI discovery. He worked on AI R&D at Google X, Nvidia and Stanford's AI Lab, has published at NeurIPS and ICML, and worked with Chelsea Finn and Andrew Ng.

What is the MLE-Bench result?

Thesis is state of the art on OpenAI's MLE-Bench, which tests how well AI systems train machine learning models autonomously. The company reached that in one month with $10k of compute, ahead of research teams at Microsoft, Google, Meta and Baidu.

Sergio wrote in 2021 about a future where AI could create its own algorithms. Foundation models made that plausible, and Thesis is the attempt to build the system rather than wait for one.

Pele Collins Luigi Charles

Sergio Charles

Co-Founder & CEO at Darwin Robotics

In this episode

Thesis

Co-Founder & CEO

Pivoted in October 2026

Now building

Darwin Robotics

Co-Founder & CEO

Robots that assemble and maintain the machines civilization runs on.

Darwin Robotics is building robots that assemble and maintain the machines civilization runs on, starting with AI infrastructure. Its robots learn complex physical tasks from demonstrations, with initial applications in AI hardware assembly and data-center operations.

Growing up in the Caribbean, Sergio taught himself calculus at 12 and general relativity at 13. By 16, he had authored research papers in differential geometry. A cold email from his home in the Bahamas led to work on an open conjecture in differential topology with guidance from John Pardon. At Stanford, he researched autonomous reinforcement learning for robotics with Chelsea Finn and computer vision with Andrew Ng. He also worked on AI R&D at NVIDIA and Google X and holds a B.S. in Mathematics and Computer Science and an M.S. in Statistics from Stanford.

Darwin (YC F25) is taking a deployment-first approach: putting robots into real workflows, identifying where they fail, and using targeted data collection and post-training to improve their capabilities. The aim is to build robots that learn new tasks quickly and perform useful work reliably across changing hardware and environments. The long-term goal is embodied artificial general intelligence.

About

Thesis is building a system that uses the outcome of each experiment to choose the next.^{[1]Deeptech DecodedSergio Charles on agents and the experiment throughput problem (20:39)Episode[2]Deeptech DecodedSergio Charles on choosing the next experiment (44:31)Episode} The point is to make machine learning research compound instead of running as a sequence of manual attempts.^{[1]Deeptech DecodedSergio Charles on agents and the experiment throughput problem (20:39)Episode} Sergio researched autonomous reinforcement learning for robotics with Chelsea Finn and computer vision with Andrew Ng at Stanford, and also worked on AI R&D at NVIDIA and Google X. He holds a B.S. in Mathematics and Computer Science and an M.S. in Statistics from Stanford. Thesis (YC F25) is state of the art on OpenAI's MLE-Bench, reached in one month on $10k of compute, ahead of research teams at Microsoft, Google, Meta and Baidu.^{[3]Launch YCThesis: Building the frontier of AI discoveryYC, Nov 11, 2025}

Co-founders

Luigi Charles

Co-founder

Episodes

How Two Brothers Outperformed Big Tech for $10K in 1 Month | (Thesis – YC F25)
Apr 30, 20261h 18m
Watch + transcript YouTube Spotify Apple Podcasts

The reasoning layer

Decoded before it's obvious.

FAQ

Sergio Charles is the co-founder and CEO of Thesis, an applied AI lab accelerating the frontier of AI discovery.^{[3]Launch YCThesis: Building the frontier of AI discoveryYC, Nov 11, 2025[4]Deeptech DecodedSergio Charles on six years in AI research and co-founding Thesis (00:40)Episode} He worked on AI R&D at Google X, Nvidia and Stanford's AI Lab, has published at NeurIPS and ICML, and worked with Chelsea Finn and Andrew Ng.^{[3]Launch YCThesis: Building the frontier of AI discoveryYC, Nov 11, 2025}

A Y Combinator F25 company building a system that runs and directs its own machine learning experiments, using each result to pick the next one.^{[3]Launch YCThesis: Building the frontier of AI discoveryYC, Nov 11, 2025[1]Deeptech DecodedSergio Charles on agents and the experiment throughput problem (20:39)Episode[2]Deeptech DecodedSergio Charles on choosing the next experiment (44:31)Episode} Its stated mission is to make state-of-the-art AI discovery 10 times faster and 10 times cheaper.^{[3]Launch YCThesis: Building the frontier of AI discoveryYC, Nov 11, 2025}

Thesis is state of the art on OpenAI's MLE-Bench, which tests how well AI systems train machine learning models autonomously.^{[3]Launch YCThesis: Building the frontier of AI discoveryYC, Nov 11, 2025} The company reached that in one month with $10k of compute, ahead of research teams at Microsoft, Google, Meta and Baidu.^{[3]Launch YCThesis: Building the frontier of AI discoveryYC, Nov 11, 2025}

Sergio wrote in 2021 about a future where AI could create its own algorithms.^{[3]Launch YCThesis: Building the frontier of AI discoveryYC, Nov 11, 2025[5]Deeptech DecodedSergio Charles on neural architecture search in 2021 (17:36)Episode} Foundation models made that plausible, and Thesis is the attempt to build the system rather than wait for one.^{[4]Deeptech DecodedSergio Charles on six years in AI research and co-founding Thesis (00:40)Episode[1]Deeptech DecodedSergio Charles on agents and the experiment throughput problem (20:39)Episode}

Fields such as medical imaging, climate, materials and robotics, and in the long run every domain of science, with the goal of a system that discovers the next Transformer or the next AlphaFold.^{[6]Deeptech DecodedSergio Charles on medical imaging, climate, materials and robotics users (26:54)Episode[3]Launch YCThesis: Building the frontier of AI discoveryYC, Nov 11, 2025}

His brother Luigi Charles, Darwin’s co-founder and CTO, who studied mathematics and computer science at Columbia and previously co-founded Sphere.^{[7]Deeptech DecodedLuigi Charles on his earlier cross-border payments startup (01:15)Episode}