Data science that ships, in whatever domain the problem lives.
Product-Focused Data Scientist Turning Models into Decisions
I build models that make high-stakes decisions in production. At Tesla, every 4680 battery cell passes through my computer vision models as the final quality gate before it goes into a vehicle. Before that, I worked on customer analytics in retail, credit and fraud risk in banking, and engine maintenance in aerospace.
Domains
Different industries, the same core work: turn messy, high-volume data into a decision someone can trust.
Lithium-ion battery reliability
Lab stress testing, CT-scan computer vision, defect detection, and thermal-runaway risk forecasting from fleet telemetry.
Shopper behavior analytics
3D pose data, in-store A/B testing, and layout simulation for physical retail.
Credit & fraud risk
Loan default modeling at scale, with explanations built for regulatory audits.
Aerospace maintenance
Engine health algorithms and geospatial tooling for fleet maintenance.
Featured work
Public write-ups of work I've contributed to, plus my patent and papers.
Engineering Tesla Vehicle Batteries for Safety, Durability and Longevity
How Tesla inspects every 4680 cell, using X-ray, CT scanning and AI-based defect detection, and how fleet telemetry feeds back into battery safety. This is the area my models and risk forecasting work in.
Read on tesla.com → Retail · Standard AIHeatmaps Are Failing Retailers. What's Next?
Why camera-by-camera heatmaps mislead retailers, and how 3D spatial reconstruction of real shopper paths gives a truer picture of in-store behavior.
Read on standard.ai → PatentSubject-tracking in a cashier-less store for spatial analytics
US 2024/0193497 A1. Co-inventor on using subject tracking in autonomous-checkout stores to improve item and shelf placement.
View patent → Paper · Under reviewAnalyzing the Shopping Journey: Computing Shelf Visits in a Physical Retail Store
Submitted to IEEE RO-MAN (34th International Conference on Robot and Human Interactive Communication). First author.
Read paper → Research · CMSAC 2021An Interpretable Method of Learning Stochastic Game Dynamics
Models a soccer ball's movement with learned potential functions to simulate matches. Captures large score differentials better than Poisson regression. Presented at the Carnegie Mellon Sports Analytics Conference.
View slides →Experience
Five roles, one thread: turning data into decisions people act on.
-
Jul 2024 – Present
Senior Data Scientist · Tesla
Own quality and risk decisions for Tesla's battery cells: the final quality gate for every 4680 cell, thermal-runaway and warranty-life forecasting, and the dashboards engineers use daily.
10+ cell programs qualified98% recall45 engineers, 4 teams200+ hrs/week saved -
Sep 2023 – Jun 2024
Data Scientist · Standard AI
Ran in-store experiments and, with product and sales, defined the shopper-engagement metrics behind a retail analytics platform.
+29% conversion$600K new contracts+15% revenue -
Jan 2023 – Sep 2023
Cell Engineer Intern · Tesla
CT-based defect detection and quality dashboards that drove battery cell design changes.
+40% defect recall−20% electrolyte leaks -
Jun 2022 – Aug 2022
Summer Associate, Credit Fraud Risk · JPMorgan Chase
Small-business loan default modeling, with risk segments explained for regulators.
+14% model lift -
Oct 2019 – May 2020
Undergraduate Researcher · Rolls-Royce
Engine-maintenance pipeline and R Shiny tool for flagging at-risk aircraft.
3–4 hrs → under 5 min
Earlier: Carnegie Mellon University (Summer Research Fellow, 2021) · LPL Financial (Information Security Intern, 2019)
Projects & competitions
Earlier side projects, mostly from UCLA DataRes and datathons.
Forecasting hot dog sales for Chevron
1st place in the Chevron track at the Rice Datathon. Lowest RMSE of 45 teams.
Customer lifetime value analysis
Projected 5- and 10-year customer lifetime value for an eCommerce company and recommended focusing on its highest spenders. Won 2nd place overall at the Retina AI Datathon.
Make-A-Monet
Image style transfer with CycleGAN and Pix2Pix, served through a Plotly Dash app.
California wildfire crisis
Led a team of six on a geospatial analysis of a decade of CAL FIRE data.
COVID-19 vaccine rollout
Prophet forecasts of vaccination milestones, plus Twitter sentiment analysis.
Variability in baseball stadiums
R Shiny app showing how ballpark dimensions affect batting outcomes.
Toolkit
Languages & ML
Data & MLOps
Analytics
Awards
- 1st Place, Chevron TrackRice University Datathon · 2021
- 2nd Place, Prosperity TrackUCSD DataHacks · 2021
- 1st Place OverallDSS Datathon for Social Good, UC Berkeley · 2020
- 3rd Place, City Search TrackTAMU Datathon · 2020
- 2nd Place OverallRetina AI R Datathon · 2020
- Gandhi Memorial ScholarshipSan Diego Indian American Society · 2019
Education
Let's talk
Based in the San Francisco Bay Area. I'm happy to hear about data science roles and interesting problems in any domain.