back to projects

~/projects/ai-simulation-platform

AI Simulation Platform

Reinforcement learning in Unity

Jan 2025 — Present·active
UnityC#PythonML-AgentsPPOPyTorch

$ overview

A modular Unity environment for training reinforcement-learning agents end-to-end — environment, reward shaping, training pipeline, and analysis.

Wrote production-grade C# for the environment side and clean Python on the training side, with TensorBoard wired in for run comparison.

$ highlights

  • Built a modular Unity ML-Agents environment with configurable reward systems and state observations.
  • Tuned PPO hyperparameters across experiments to improve learning stability.
  • Production-style C# integrated cleanly with Python training pipelines.