Tom Rossmann

Applied Data Scientist

I’m studying for a master’s in mathematics and working on machine-learning projects, from time-series forecasting to agents that learn through play.

Projects

An agent that learns through self-play to make long-term decisions in a dice game with explicit chance nodes.

81.1% win rate against a heuristic baseline over 512 games

JAX · Flax · RLax · MCTS · self-play

Multi-horizon solar irradiance forecasting across short and long time ranges, evaluated against seasonal and persistence baselines.

GRU reduced RMSE by about 34% at 4h and 21% at 24h against the strongest naive baseline

GRU · LSTM · time series · baselines

A comparison of ridge and kernel ridge regression, including an algebraic proof of their equivalence.