Jeremy Gu

Data Science & AI Leader | Educator | Open Source Contributor
San Francisco Bay Area

I build production data and ML systems, lead the teams that ship them, and teach applied data science. My current work spans LLM applications, RAG, coding agents, evaluation workflows, and the reliability layer around AI systems.

Leadership · Education · Teaching · Open Source · LinkedIn · Email

• SageSure — Director of Data Science (2025): Led AI and data products for underwriting, claims, support, RAG, computer-vision triage, and human-in-the-loop evaluation
• Shipt — Director of Data Science (2021–2024): Built and led a 17-person DS/ML team across marketplace optimization, ETA, fraud and risk, experimentation, and retail analytics
• Stitch Fix — Data Science Manager (2019–2021): Led channel optimization, marketing measurement, incrementality, LTV, and value-based bidding
• Uber — Tech Lead Manager, Strategic Finance (2017–2019): Led finance data science for forecasting, planning, growth analytics, and executive decision support
• Amazon Web Services — Applied Scientist (2014–2017): Built ML and analytics systems for marketing, sales, recommendations, ads, and customer-growth decisions

• Stanford University — Graduate School of Business and CS Graduate Studies, 2024–2026; 4.0/4.0 GPA in computer science coursework
• University of Washington — MS in Statistics
• University of Minnesota — BS in Mathematics and Statistics

I teach applied data and AI courses that connect technical foundations with production practice.

Public course materials

• MSDS 682: Data Stream Processing — Summer 2026 · University of San Francisco · Course repository

Other courses taught

• BUS 36 · Summer 2024 and Spring 2025

Additional public course materials will be added as they are released.

Selected from 16 merged upstream pull requests by wgu9, ordered by impact:

• LightGBM #7348 — Made incomplete binary-file writes fail safely in C++
• MLX #4086 — Fixed strict lexicographic ordering for Device and Stream
• XGBoost #12280 — Included the license file in Python wheel metadata
• segment-geospatial #538 — Added SAM 3.1 model support
• MLX #3748 — Added vecdot to the Array API namespace
• OpenHands #14810 — Made trajectory exports explicitly UTF-8 with regression coverage
• scikit-learn #34317 — Made Array API metric references explicit

Authorship credit: NousResearch/hermes-agent #51678 includes maintainer-merged work derived from my contribution.

Current work in review spans AI agents, model serving, ML frameworks, retrieval evaluation, and systems reliability. See open public pull requests.