Projects

Robotics

A general harness that connects coding agents to physical robots, simulators, and models.

Agentic reinforcement learning

STACX

STandAloneCompleX: modular infrastructure for end-to-end agentic reinforcement learning.

Machine learning agents

MLE-Dojo

Interactive environments for developing LLM agents that perform machine learning engineering tasks.

Generative models & RL

Flow RL

A high-performance reinforcement learning library built around flow and diffusion models.

Representation learning

Representation-based Reinforcement Learning

Research implementations of reinforcement learning algorithms for stochastic nonlinear dynamics through learned representations.

Control

Representation-based Control Toolbox

A solver for nonlinear stochastic control using spectral representations.