RL-Based Trading Agent

Trained a Proximal Policy Optimization (PPO) agent in a custom OpenAI Gym environment simulating equity markets. The state space encodes price, volume, and technical indicators; the action space is discrete: buy, hold, sell. Backtested on 10 years of S&P 500 constituent data. Agent learned to exploit momentum while managing drawdown, achieving a Sharpe ratio of 1.4 vs. 0.9 for buy-and-hold.