CS, Stats @ Stanford

Diego Sánchez

Diego Sánchez

I'm a computer science and statistics student at Stanford interested in ML systems, inference, and the infrastructure that makes machine learning reliable in practice. I especially enjoy problems around routing, experimentation, performance, and understanding how systems behave as workloads and assumptions change.

Beyond ML systems, I care about how technology can create meaningful impact in healthcare, education, and accessibility. I'm particularly interested in building tools that are useful beyond the people technology traditionally serves, an interest shaped in part by growing up across Miami, Oaxaca, Chicago, and California.

Previously, I've worked on adaptive ML infrastructure at Coinbase, LLM search and evaluation at Microsoft, and healthcare-focused LLM systems at Stanford Medicine. Across those experiences, I've gravitated toward work that sits between research and engineering: taking an idea, understanding how it behaves, and making it work inside a real system.

How the Serving Lab works

Experience

  • Coinbase

    Research Engineer Intern

    June - Sep 2026

    San Francisco, CA

    Built a Bayesian ML rollout policy for adaptive promotion decisions, replacing a fixed four-day ramp and running inference inside the serving budget.

    0.36 ms p50 · 200+ live gRPC calls · zero routing errors

    Technical detail

    Identified a 19.7% false-full-ramp failure mode in a production mSPRT comparison and designed a guard reducing it to 10.4% while retaining 94% power.

  • Microsoft

    Software Engineer Intern · M365 Copilot Search NLP

    Jun – Sep 2025

    Redmond, WA

    Built an LLM pipeline for triaging M365 Copilot Search feedback, combining model evaluation, human review, and privacy-aware processing.

    97% recall · 92% precision · 99%+ PII removal

    Technical detail

    Reduced estimated yearly processing cost from ~$430k to ~$4.2k by automating triage workflows.

    Also at Microsoft: 3rd place globally · Microsoft Intern Hackathon

  • Stanford Medicine

    LLM Developer Research Assistant · SURF Stanford Medicine

    Jan – Jun 2025

    Stanford, CA

    Built a HIPAA-conscious FastAPI backend and LLM system for real-time unit analytics and predictive forecasting, with automated reference-response evaluation.

    95% reported LLM output accuracy

Programs & Community

Selected Work

A couple of systems projects I keep coming back to.

More on GitHub .

Serving Lab

Open this page on a desktop to try the interactive Serving Lab.

How this works

Miscellaneous :)

Stanford University: B.S. Computer Science (AI Track) and Statistics, Class of 2028.

I'm a first-generation Oaxacan-American. I was born in Miami, spent much of my childhood in Oaxaca, lived in Chicago, and eventually made my way to California.

I've played trumpet for more than 10 years and still love recitals, orchestras, and live music.

I love playing and watching football. I'm a Real Madrid fan, and three of my favorite players are Luka Modrić, Cristiano Ronaldo, and Park Ji-sung.

I'm usually reading magical realism, watching anime, reading manga, or finding some way to help someone out.

Want to talk about ML systems, infrastructure, or something interesting?

dsanh14@stanford.edu