Selected Work & Case Studies
03 Projects
Predicts F1 fantasy points using an XGBoost pipeline trained on historical race and practice session data, then solves optimal team selection as an integer programme under budget, roster, and transfer constraints. Results are presented through an interactive Streamlit dashboard.
Trains neural networks to play Flappy Bird using NEAT, evolving topology and weights from minimal structure. Includes a browser-based replay viewer for stepping through generations.
Analyses 30-second rainforest recordings to produce a 0–100 ecosystem health score, combining acoustic indices (ACI, BI, NDSI) with a PyTorch model trained on the Rainforest Connection dataset.