The toolbox
What I use
The languages, libraries, and tools I reach for.
Languages
Python for almost everything, C++ and C when speed matters, SQL for data.
Machine learning
PyTorch, Stable-Baselines3, and Gymnasium for reinforcement learning; Transformers and sentence-transformers for NLP and RAG; scikit-learn, NumPy, and Pandas for the classic stack.
Scientific computing
Numba for JIT-compiled physics cores, and numerical methods (RK4, DMD, FDTD) when a problem calls for them.
Visualization & tooling
Matplotlib for plots, Three.js / WebGL for interactive views, Git for version control, and this Astro site for writing it all down.