Jazmia Henry Building worlds that train more capable models

I'm a foundation model engineer: at Collide I own post-training end to end (SFT, DPO, GRPO), reward modeling, and the RL environments behind models from 8B to a 120B MoE. Before that: production RL agents at Microsoft for Shell, PepsiCo, and Abbott; Head of Machine Learning at The Motley Fool; founder of Iso AI, where my simulation-based fine-tuning method is patent pending. NeurIPS author (Top 5% paper) and reviewer, Stanford HAI affiliated fellow, LinkedIn Learning instructor. I love logic, game theory, and teaching hard things plainly. I know a little bit about a lot, and I've gone deepest on how models learn.