An inverse physics-informed neural network that recovers a material's thermal diffusivity from sparse, noisy temperature measurements, benchmarked against classical parameter fitting.

Three small simulations — a random walk, radioactive decay, and a toy particle-physics bump hunt — each checked against the analytical result it is supposed to reproduce.

A dependency-free Python converter that turns plain-text multiple-choice questions into Perseus JSON, with validation that refuses to emit a broken file and a report on answer-key bias.

A self-directed curriculum working from mathematical foundations toward physics-informed machine learning, implemented from scratch.