projects
2025 — adaptive reinforcement learning ensemble strategy, queen's ai club
> implemented and trained an ensemble of ppo, a2c, and td3 reinforcement learning agents, with experiment tracking and version control managed via git, enabling adaptive allocation across multiple market regimes.
> engineered a diversification framework that reduced portfolio volatility and improved risk-adjusted returns, validated by sharpe ratio and drawdown analysis versus spy buy-and-hold.
2025 — autonomous multi-agent robotic firefighting
> developed and optimized lloyd's algorithm in matlab for adaptive k-means clustering of dynamic fire hotspot data, reducing cluster convergence time by 30% and improving autonomous robot deployment efficiency.
> leveraged gis wildfire spatial data and analysis tools to delineate and model fire perimeters, enabling accurate real-time input for robotic firefighting cluster optimization and enhancing fire containment strategies.