P01Identities Data Science Tech Lead
LLM classification at Fortune-500 scale
100K+ records auto-classified
Designed and shipped an LLM-powered pipeline running Claude and GPT inside Snowflake Cortex to classify business types and standardize merchant addresses across 100K+ records. Replaced a manual review queue that had been the team's biggest scaling bottleneck. Result: classification latency dropped from days to minutes, accuracy held above the human baseline, and downstream risk models gained a cleaner upstream signal.
- Snowflake Cortex
- Claude
- GPT-4
- Python
- Airflow
P02Identities Data Science Tech Lead
$0.5M/yr fee-exposure classification model
$500K annual risk uncovered
Built a classification model that surfaced previously invisible fee-exposure patterns across the merchant base, translating to roughly $0.5M of annualized risk. Paired the model with rigorous causal evaluation — difference-in-differences, segmented regression, and survival analysis — so leadership could act on the finding with confidence.
- Python
- scikit-learn
- diff-in-diff
- Survival analysis
- Snowflake
P03Identities Data Science Tech Lead
GooseAI for CS-ticket churn analysis
Churn drivers identified at scale
Top user of Square/Block's internal GooseAI LLM. Used it alongside ChatGPT to mine thousands of support tickets, surfacing the failure patterns and friction points that correlated most strongly with churn. Findings fed directly into product prioritization for the following quarter.
- GooseAI (Square/Block internal LLM)
- ChatGPT
- Python
- SQL
P04Senior Data Scientist · E-Commerce & Mobile
Org-wide A/B testing rollout
Optimizely + office hours
Led the Optimizely rollout for the e-commerce org and co-hosted recurring company-wide A/B testing office hours. Designed and analyzed experiments across onboarding, the website editor, and OS deprecation — equipping non-DS partners to run statistically defensible tests themselves.
- Optimizely
- Looker
- Causal inference
- Statistics
P05Identities Data Science Tech Lead
ETL template + data contracts adoption
50% time saved on new pipelines
Authored the org-wide ETL template and championed adoption of Anomalo, Select Star, Growthbook, and data contracts. The combination cut new-pipeline build time roughly in half and gave the team early-warning signals for data quality regressions. Also served on the AI Steering Committee shaping the enterprise AI roadmap.
- Airflow
- Snowflake
- Anomalo
- Select Star
- Growthbook
P06Community leadership
Data Town Square
200+ practitioners convened
Co-led a virtual Data Town Square that grew to 200+ attendees — a recurring forum for practitioners to share patterns, anti-patterns, and frank technical critique across teams and companies.
- Public speaking
- Community
- Cross-team facilitation
