Data About Dave

Because what kind of data scientist would I be if I didn't quantify myself?

View My Portfolio Explore AutoTrader

Career By The Numbers

0 Years Writing Code Professionally since age 16
0 Companies Biotech to fintech
$0 Estimated Business Impact $20M documented ยท $150-300M+ incl. Shipt GMV lift, Riviera 4x ARR, Freenome product & platform
0 Models Trained AutoTrader alone
0 Max GMV Lift Shipt A/B-tested shelves
0 Scientists Supported Freenome ML platform
0 Engineered Features AutoTrader feature store
0 Berkeley Degrees CS + Data Science

What I Talk About Most

Keywords extracted from my resume, portfolio, and project documentation. Sized by frequency. Hover for context.

Primary language across every rolePython Title, identity, and passionMachine Learning Current focus at ShiptPersonalization Shipt two-tower + AutoTrader featuresEmbeddings Shipt + AutoTrader + FreenomeFeature Engineering Built infrastructure at every companyInfrastructure Core at Shipt, Riviera, AutoTraderRecommendations Freenome + AutoTrader + Shipt GCSGCP Used at AutoTrader, Shipt, Riviera, CastlightXGBoost Freenome platform + Shipt experimentsA/B Testing Used across every roleSQL Approximate nearest neighbors at scaleFAISS Freenome deep learning adoptionPyTorch Shipt two-tower recommenderTwo-Tower Core ML task across careerClassification AutoTrader + Shipt + FreenomePostgreSQL Sentic package + Doximity + AutoTraderNLP Shipt real-time recommendation deliveryKafka Riviera, Castlight, Change, AutoTraderRegression Freenome experiment trackingMLFlow Taught 50+ people at FreenomeGit Freenome + Shipt orchestrationAirflow Freenome core productCancer Detection Always ships to productionProduction Systems Freenome + AutoTraderDeep Learning Shipt GMV, engagement, NDCGMetrics Design Shipt RAG + agentic frameworkRAG Riviera + Shipt shelf rankingRanking Models AutoTrader hyperparameter searchOptuna Shipt data warehouseSnowflake Freenome ML executionFlyte Change Healthcare cloud toolsAWS Freenome containerizationDocker Freenome orchestrationKubernetes Shipt Interaction ScoreComposite Scoring Shipt + Freenome establishedCI/CD EEG research at BerkeleyBrain-Computer Interface Sentic package, 20+ languagesSentiment Analysis AutoTrader walk-forward validationTime Series Freenome petabyte-scale genomicsGenomics AutoTrader VWAP labelsVWAP

A Few More Charts

Skills Depth

Self-assessed relative proficiency across disciplines

Time Across Domains

Approximate years spent in each industry

Top Keywords by Frequency

How often key terms appear across my resume, self-evaluation, and project docs

Technology Timeline

When I picked up key technologies and how long I've used them professionally

2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025

Things That Don't Fit on a Resume

The Running Club
I built an ML model to predict depression from lifestyle data. Its feature importances showed me that exercise and social activity mattered more than I thought. I joined a running club the next week and haven't stopped.
$55/Month
AutoTrader trains 1,800+ models, generates predictions for 600+ tickers, and delivers subscriber emails every morning—all on two GCP VMs that cost less than a gym membership.
The Git Teacher
At Freenome I taught git to over 50 scientists and researchers through an internal instructional series. Turns out the hardest part isn't rebasing—it's convincing a PhD that version control is not optional.
16 Repositories, 1 Person
When Shipt's Personalization team was first formed, I was the only data scientist for four months. Maintained all 16 Discovery Science repositories solo while simultaneously designing the next generation of infrastructure.
The Honest Metric
At Shipt, I discovered that the team's flagship KR (5% GMV attribution) couldn't be reproduced. Instead of quietly working around it, I documented the discrepancy and presented the real number (1–4%). It led to better metrics for everyone.
XRP to S&P 500
My first algorithmic trading project was a Ripple (XRP) sentiment bot in 2014 that won 3rd place in a Ripple API contest. A decade later, I'm running 1,800+ models across the S&P 500. The itch never went away.