Built a machine learning pipeline to map cancer progression trajectories from whole-slide histopathology images. Combines deep learning tissue embeddings (Phikon) with diffusion pseudotime analysis to construct a spatial reference atlas showing how breast cancer morphology evolves within and across patient samples.

Implemented and benchmarked RL models (Q-Learning, Double Q, Choice Kernel, Forgetting) for probabilistic choice in mice, with a Python pipeline for cleaning, simulation, and synthetic datasets, plus hyperparameter tuning to improve fit quality and parameter recovery (Double Q + Forgetting strongest).