Building Thesis, an applied AI lab accelerating the frontier of AI discovery. The platform lets researchers run experiments, train models, and discover what matters — entirely in silico, autonomously, and at scale. Previously worked on AI R&D at Google X, Nvidia, and Stanford's AI Lab (SAIL), with published work at NeurIPS and ICML and research alongside Chelsea Finn and Andrew Ng. Holds a B.S. in Math & Computer Science and an M.S. in Statistics from Stanford. Across those roles, Sergio kept hitting the same gap: researchers building models had no real tooling. Thesis (YC F25) is the answer — state-of-the-art on OpenAI's MLE-Bench in just one month and $10k of compute, and already trusted by researchers at Stanford, Harvard, MIT, and Mayo Clinic.
Episode featuring Sergio Charles will be out soon.
Thesis is an applied AI lab accelerating the frontier of AI discovery. The platform lets researchers run experiments, train models, and discover what matters — entirely in silico, autonomously, and at scale — across architectures like LSTMs, Transformers, GRU Seq2Seq, Wavenet, and Temporal Fusion Transformers.
Sergio kept hitting the same gap across Google X, Nvidia, and Stanford's AI Lab: researchers exploring data, running experiments, and building models had no real tooling. Thesis is built specifically for AI R&D — treating autonomous experimentation and model training as a first-class workflow rather than stitching together notebooks, schedulers, and one-off scripts.
MLE-Bench is OpenAI's benchmark for how well AI systems can train ML models autonomously — a first step toward self-improving systems. Thesis reached state-of-the-art on MLE-Bench in just one month of work and only $10k of compute, outperforming teams of researchers at Microsoft, Google, Meta, and Baidu.
Sergio Charles (CEO) and Luigi Charles (CTO) — brothers. Sergio previously worked on AI R&D at Google X, Nvidia, and Stanford's AI Lab, with published work at NeurIPS and ICML and research alongside Chelsea Finn and Andrew Ng. Luigi previously built Sphere, a cross-border payments company that processes billions of dollars annually and was valued at $250M+ in its Series A.
Researchers at Stanford University, Harvard University, MIT, and Mayo Clinic, among others. Thesis is backed by Y Combinator (F25 batch) and ships a native macOS application.
δ (delta) is the mathematical symbol for change — a nod to the company's mission of compounding scientific progress. Thesis publicly launched as 'δ Thesis: Building the frontier of AI discovery' through Launch YC.
To make it possible for a researcher anywhere to discover the next Transformer or invent the next AlphaFold. The team is building toward autonomous, self-improving AI research systems — compressing experimentation cycles so breakthroughs arrive faster and at a fraction of today's cost.
