Machine Learning Projects

Summary

Details of these projects are under NDA and cannot be publicly disclosed. Across my time at Unity Technologies and Capgemini, I contributed to the following machine learning initiatives.

• Reinforcement Learning for Unitree H1 (Sim2Real)

  Trained a reinforcement learning model for the Unitree H1 humanoid robot in simulation, targeting Sim2Real transfer to the physical robot. Built simulation environments in NVIDIA Isaac and MuJoCo, writing the core training logic in Python with CUDA-accelerated training to teach the robot to walk and stand.

  • Reinforcement learning training pipeline for the Unitree H1 humanoid robot
  • Simulation environments built in NVIDIA Isaac and MuJoCo
  • Sim2Real transfer enabling deployment to the physical robot
  • CUDA-accelerated training logic implemented in Python

• Synthetic Data Generation for Perception ML

  Built a Unity-based pipeline to generate tagged synthetic 3D data, used to train a perception ML model to identify different objects in the real world.

  • Unity-based synthetic 3D data generation pipeline
  • Automated tagging to support supervised model training
  • Trained a perception ML model to recognize real-world objects