Aryan Turalkar

Multi-Agent DRL Portfolio Optimization: Implemented a PPO actor-critic policy with modality-specific encoders (price LSTM, sentiment FFN, volume FFN), custom portfolio environment with Differential Sharpe Ratio reward shaping and GAE-based advantage estimation, trained on NSE data. Shipped a FastAPI backend and a React dashboard showing per-stock action traces and backtest results. Ran ablation studies across input modes to identify the best configuration.

Residential Governance & Analytics Portal: Built interfaces and workflows for housing society operations including notices, maintenance, and user activities. Designed authentication and role-based access flows for different user types.

Built a reservation management system for campus resources. Designed backend APIs and database schemas for booking operations. Containerized application using Docker and automated test and deployment pipelines via GitHub Actions CI/CD.