About
Builder, scientist, leader.
I lead engineering at Stanford University for ROAR, a K-12 reading-assessment platform deployed in 309 districts and 2,700+ schools, used by 160K+ students. I built the engineering team and own the technical decisions that take ROAR from a research platform to a system districts can trust for classroom use. Hiring, roadmap, architecture, vendor management, FERPA-grade data governance, budget. All of it lives with me.
AI is changing what a small engineering team can do. We use coding agents, AI code review on every PR, and generative testing on the code paths where mistakes would matter most. The work that decides whether any of this helps isn't the prompts. It's the documentation underneath: ~70 pages of domain knowledge, project conventions, and review heuristics that ground every AI-assisted change. Inside the product, the focus is psychometric integrity: validated assessments built on modern measurement theory (CAT, latent-ability models, IRT item banks), with validation work done in partnership with the Stanford Accelerator for Learning and published in peer-reviewed journals. Both halves of that get harder when the users are children.
Before this role I was a postdoc at the UW eScience Institute, where I built open-source tools that ended up in research workflows: pyAFQ (Nature Methods), Cloudknot (SciPy), Groupyr (JOSS), and AFQ-Insight (PLOS Computational Biology). My PhD is in computational physics from the University of Washington, where I ran Monte Carlo simulations on DOE supercomputers as a Computational Science Graduate Fellow and published in Physical Review Letters.
Before grad school I served as a U.S. Peace Corps health-education volunteer in Morocco, designing federally funded youth programs in rural schools and co-organizing the country's first national English spelling bee. Before that I was a U.S. Air Force officer and systems engineer at the Space and Missile Systems Center, where I program-managed an $8M software platform and helped link a $2.4B lab budget with a $12B satellite R&D portfolio.
The thread through all of it is an unusual mix: scientific rigor, structured engineering, and the slow work of building community. All three show up in what I lead now.
What I Own
- ▸ Engineering strategy, technical roadmap, and operating cadence
- ▸ Hiring, mentorship, and team structure (currently six engineers)
- ▸ AI-accelerated engineering practice — tooling, QA, evaluation
- ▸ End-to-end architecture: web platform, APIs, data pipelines, integrations
- ▸ Reliability discipline: monitoring, incident response, release management
- ▸ Data governance, privacy & user safety on a K-12 platform (FERPA-aligned)
- ▸ Technology budget, vendor selection, and contractor management
- ▸ Cross-functional partnership with Product, Research, and 25+ school districts
How I Work
AI is only as useful as the documentation underneath it
On a small team, AI either compounds your output or wastes your time. The work that decides which way it goes isn't the prompts. It's the domain knowledge, conventions, and review heuristics that any engineer or coding agent can learn from before writing a line. We've written ~70 pages of that for our dashboard repo alone.
80% of the value with 20% of the effort
Building for the future isn't the same as over-engineering for the present. I push the team to find the smallest version of the right thing and ship it, then earn the right to iterate. Most of the wins on the platforms I've led came from saying no to the overbuild.
Reliability is the cheapest feature you can ship
Monitoring, audit logs, CI/CD, and a clean on-call rotation don't slow a team down. They're how you avoid spending the next quarter fighting the fires you should have caught. The setup work pays back fast.
Engineering and Product is a partnership
The teams I've been part of that worked best had clear partnership between an engineering leader and a product leader. The engineering side owns how. Product owns what and why. I prefer to share accountability for both, and work disagreements out before either of us walks into a room of stakeholders.
Privacy and safety are design constraints, not paperwork
Working on a platform that serves children raises the floor on what safety and privacy mean. Least-privilege access, system-of-record clarity, retention and deletion handled in code, vendor review. These aren't compliance theater. They're how you avoid being the headline.
Timeline
2023–present
Director of Technology & Innovation
Stanford University — ROAR / Graduate School of Education
Leading technical development of web-based reading assessments and data infrastructure.
2022–2023
Research Scientist
Stanford University — ROAR / Developmental-Behavioral Pediatrics
Data science, neuroimaging, and educational technology.
2020–2022
Data Science Postdoctoral Fellow
University of Washington — eScience Institute
Built open-source tools for reproducible neuroimaging analysis.
2014–2020
PhD, Nuclear Physics
University of Washington
Computational Science Graduate Fellowship. Quantum Monte Carlo simulations on DOE supercomputers.
2011–2013
Peace Corps Volunteer
U.S. Peace Corps — Kingdom of Morocco
Health education, community infrastructure, and peer education programs.
2006–2010
Air Force Officer & Systems Engineer
U.S. Air Force — Space and Missile Systems Center
2010
MS, Condensed Matter Physics
California State University, Long Beach
Graduate Dean's List (top 1.5%).
2006
BS, Engineering Physics
Embry-Riddle Aeronautical University
Summa Cum Laude.