Lean & AI for mathematics
Formal verification, proof automation, and machine-checked reasoning with the Lean theorem prover. We help teams build systems that do — and verify — real mathematics.
Boutique AI research studio
Conic Lab is a small, senior research studio. We work hands-on with a select group of teams on the most demanding parts of their AI roadmap — with a particular focus on Lean and AI for mathematics.
Led by Fei Xie, a mathematician-turned-ML practitioner working at the frontier of AI for mathematics and LLM agents.
About Fei
A research partner for teams pushing the limits of what models can do.
Conic Lab is a niche research and engineering studio working across the modern AI stack — from the safety of autonomous agents, to the infrastructure that trains large models, to formal reasoning and automated mathematics.
We deliberately stay small and take on a handful of engagements at a time. That means senior people doing the actual work, deep collaboration with your team, and a bias toward the problems that are too hard or too specialized for a general-purpose vendor.
Most of our work is confidential. What we can say is that it tends to live at the research frontier — where the right answer isn't known yet.
Formal verification, proof automation, and machine-checked reasoning with the Lean theorem prover. We help teams build systems that do — and verify — real mathematics.
Guardrails, evaluation, and safety mechanisms for LLM agents operating in the wild.
The systems and tooling that make large-model training fast, stable, and reproducible.
Turning frontier ideas into working systems — prototyping, evaluation, and the messy path to production.
An AI-for-mathematics effort built on the Lean theorem prover.
Lean · MathSafety features for LLM agents.
Agent safetyInfrastructure for large language model training.
Training infraEngagement details are confidential. We share only what our clients have made public.
Have a problem that doesn't have a known answer yet? Let's talk.