Data science that ships, in whatever domain the problem lives.

Senior Data Scientist

• Run lab stress tests on vendor and in-house battery cells under 98th-percentile fleet conditions, then combine the results with telemetry from 3.5 million vehicles to forecast thermal-runaway risk. These forecasts qualify next-generation batteries and inform go/no-go decisions with leadership.
• Built and deployed U-Net models that segment 50+ battery components from CT scans, reaching 98% recall on critical checks. They are the final quality gate for every Tesla 4680 cell.
• Built a Celery/Redis ETL pipeline on 6 VMs that runs CT inference in 2 minutes per cell (90 scans total) across 2,000+ batteries a day, logs metrology (electrode overhang, core concavity, cathode gap) to a central database, and triggers Splunk alerts on model-flagged out-of-spec cells. It feeds real-time quality dashboards and saves 200+ technician hours a week.

Data Scientist

• Designed and led in-store A/B tests for Compass Group USA on product placement near store exits, lifting interaction-to-purchase conversion from 24% to 31%.
• Developed a Python shelf-interaction library on 3D pose data and store maps in Databricks and Snowflake. It powers our shopper-analytics platform (engagement funnels, cohorts, repeat visits), which won $600K in new contracts within 3 months.
• Built a Monte Carlo engine simulating shopper paths between shelves to test layout swaps before changing stores (e.g., swapping back-wall energy drinks with entrance snacks). The recommended swaps lifted revenue by 15% across two stores.

Cell Engineer Intern

• Built a Vision Transformer that detects core impingement and collapse in CT scans, improving defect-detection recall by 40% over the previous binary filter.
• Replaced several $27K Keyence machines with a CT-based method that non-invasively measures terminal/lid bulge on cells to within 50 µm.
• Built a daily-refreshing Tableau dashboard linking dissection reports to production metrics. It surfaced failure trends that drove electrolyte loading and cell design changes, cutting electrolyte leaks by 20% in 5 months.

Summer Associate, Credit Fraud Risk

• Improved the PySpark small-business loan default model by 14% with features for erratic repayment, seasonal revenue, and spikes in credit utilization.
• Made high-risk applicant segments explainable for federal regulatory audits using feature binning and hierarchical clustering.

Undergraduate Researcher

• Built an algorithm for the optimal compressor wash interval on AE3007 engines and shipped it as an R Shiny app, cutting analysis time from 3–4 hours to under 5 minutes. It mapped and isolated hundreds of at-risk aircraft.

Research & internships