I study and build reliable machine learning systems — with interests spanning learning theory & optimization, evaluation/calibration, conditional compute, and scientific ML. I like problems where rigor meets engineering: clean objectives, tight measurements, and runnable artifacts.
- 🎓 Stanford’27, CS (AI track) — coursework in ML theory, CV/DL, NLP, self-improving agents, linear & convex systems.
- 🧪 Research: Microsoft Research (AI Interaction & Learning), Stanford Scaling Intelligence Lab; prior applied ML in drug discovery and edge/streaming systems.
- 🔬 Interests: controllable/model editing, calibration/uncertainty, safe RL, and adaptive inference.
- Evaluation for agentic systems — policy families, ablations, and uncertainty-aware scoring with expert-in-the-loop review.
- Controllable model editing — task-vector approaches to adjust behavior trade-offs (fairness/alignment vs. accuracy).
- Attribution drift & OOD — studying shifts in saliency/attention as early signals for distribution change.
Toolbox: PyTorch · Hugging Face · ONNX Runtime · scikit-learn · Python · (some) C/C++ · JS/TS
Email: lpgomez [at] stanford.edu • LinkedIn: linkedin.com/in/laura-gomezjurado