Projects

Developed a numerical solver for Hamilton-Jacobi-Bellman (HJB) equations used in optimal control problems. The solver is based on finite difference methods and semi-Lagrangian schemes to handle nonlinear dynamics.

Designed and trained machine learning models to predict Walmart sales using historical data. Used time-series decomposition, feature engineering, and XGBoost/LightGBM to improve accuracy on hierarchical datasets.

Designed adaptive path-planning strategies for autonomous swamp robots navigating dynamic terrains. Implemented obstacle avoidance using sensor feedback and control-theoretic approaches.