Overall

I train and evaluate agentic AI, currently focused on adversarial robustness. Before that I worked on ranking and retrieval models in Google Search, and on training and evaluating the LLMs behind AI Overviews and AI Mode.

Work Experience

Senior Research Engineer @ Google DeepMind

Nov 2025 – Present
  • Post-training of Gemini, focused on adversarial robustness. Built and maintain an internal framework for automated red teaming, now used by teams across several product areas.
  • Drove two evaluation RFCs from proposal to acceptance; both now run against every Gemini release candidate.

Senior Software Engineer @ Google Search

Jul 2022 – Nov 2025
  • Pretrained early versions of the LLM behind “AI Overviews”, improving internal evaluations for international languages by 50%+ after tracing the cause to the production tokenizer.
  • Sole author of the post-hoc attribution pipeline that grounds fine-tuning data for Search LLMs — still the workhorse for that work three years on.
  • Engineered a regulation-compliant, petabyte-scale data store feeding the most active ML models in Search.
  • Automated data generation, training, and evaluation for a critical ranking model, cutting refresh cycles from 6–12 months to monthly.

Process Engineer @ Lam Research

Jan 2020 – Jun 2022
  • Invented and patented a physics-based model for predicting EUV lithography patterning (paper), cutting 6+ hour simulations to seconds. 10+ related patents.
  • Founded and led the image analysis package that brought that model into production use.

Education

Ph.D. in Chemistry

Aug 2014 – Dec 2019
Massachusetts Institute of Technology

Semiconductor and magnetic nanocrystal research — Google Scholar.

Research from before my time earned my advisor the 2023 Nobel Prize in Chemistry.

B.S. in Chemistry & Mathematics

2010 – 2014
University of Arizona

Skills & Clearances

Languages

Python R C++ R SQL Bash

* R indicates formal Google “Readability” credential.

ML & Research

JAX Flax PyTorch TPUs Reinforcement learning Post-training Automated red teaming Eval design for agentic systems

Infrastructure

Apache Beam Borg