Welcome to my feed

Leduc poker solver using counterfactual regret minimization. Built in C++17 with full training, playable solver, and test suite.

A platform that lets any AI agent talk with users over a phone call by handling real-time speech recognition, AI processing, and natural voice responses end-to-end. Winner of AI ATL 2025 Drive Capital track.

Autonomous vehicle navigation using deep reinforcement learning. Built with highway-env and stable baselines3, using PPO with custom reward shaping.