约翰·普拉特John Platt

Wrote two of the pieces of classical machine learning that almost everyone still uses without naming them: SMO (1998), which made training support vector machines tractable, and Platt scaling (1999), which turns a classifier's raw scores into calibrated probabilities. Spent years at Microsoft Research before joining Google, where he now leads Applied Science — computer science pointed at physical and biological problems, with climate as the main target. Won a Technical Academy Award in 2006 for computer graphics work, and has two named asteroids. Recent work includes an agentic research system that grew out of an attempt to automate Kaggle and combines LLMs with tree search to improve experiments.