Jordan Taylor

Mechanistic interpretability, mechanistic anomaly detection, model internals techniques and AI safety techniques generally.