I'm a robotics & controls engineer interested in:
- Reinforcement learning
- Physics simulation and world models
- Control theory
- Optimization
By day, I work on autonomous vehicles at Kodiak AI. By night, I maintain TorchRL, PyTorch's official reinforcement learning library.
Here's an example of what I do: look at these driverless trucks running my code, operating autonomously 24/7 with nobody in the cab!
Read more about these trucks here.
- AMSwarm 2.0: An ultra-fast drone swarm trajectory planning algorithm written in C++ with Python bindings
- 🥉 3rd place in Mistral x Hugging Face Robotics Hackathon 2025 with a bimanual robotic system that interprets natural language instruction and packs your bag for you:
They say the best way to learn software is by contributing to open-source projects. Here are some contributions I'm proud of:
- State-of-the-art RL algorithm implementations (Link)
- A more advanced and flexible optimization API for the RL trainer ecosystem (Link)
- Auto-configuration of exploration module specs from RL environments (Link)
- Various small improvements, such as:
- MPS compatibility fixes e.g. 1, 2
- config classes, e.g. TanhModule and TensorDictSequential
- BASc Mechanical Engineering @ University of Toronto
- MSc Robotics, Systems & Control @ ETH Zurich