Computer Engineering at McGill. I work on reinforcement learning and control for robots: humanoids at Mila / McGill Mobile Robotics Lab, and the perception stack of McGill Robotics' autonomous underwater vehicle.
🌐 kaidalisohaib.github.io · 💼 LinkedIn · ✉️ kaidalisohaib@gmail.com
| Humanoid balances a pole while walking | Friction-Cliff-Bench | One policy, five robots |
|---|---|---|
![]() |
![]() |
![]() |
| 31-DOF whole-body RL controller for the Unitree H2, using only its head camera. 99.0% success walking | GPU vehicle sim (NVIDIA Warp) + C++/CUDA MPPI, benchmarked against PPO on 3,009 paired scenarios | One PPO policy co-trained across five biped robots. Gait asymmetry 17.5% → 4.0% |
Other work
- 🌊 McGill Robotics AUV: real-time C++ / ROS 2 perception on a Jetson Orin (TensorRT detection + Kalman tracking at 30 Hz); 6th overall at RoboSub
- 🤖 Sumo-Bot: autonomous combat robot on an ESP32 with FreeRTOS
- 📍 AMCL-Visualization: particle-filter localization in Unity
Tools: Python · C++ · CUDA · PyTorch · Isaac Lab · MuJoCo · ROS 2 · NVIDIA Warp · Docker






