Data Analyst · Python · SQL · Tableau
I'm Yougi — an Informatics & Psychology student at UMass Amherst building dashboards, models, and data pipelines, including a live NYC collisions app over 27K records.
A public Streamlit app analyzing 27,164 NYC collisions (Jan 2020 – May 2025) — 13,508 injuries and 64 fatalities across all five boroughs. Four SQL analyses (severity, borough, hour-of-day, monthly trends) sit on a validated 23-field schema, with caching and pytest/CI for fast, reliable loads.
A scheduled GitHub Actions pipeline that samples one domain × technique pairing each weekday from an 89-cell matrix and has the Claude API draft a small, standalone analysis project for it. A manual promote/reject gate feeds back into sampling weights — decay-weighted, Laplace-smoothed promotion rates with a novelty bonus — so exploration never shuts off.
Turns a 64,000-customer randomized email A/B test into a targeting strategy. The Women's email's +4.5pp average visit lift is really +7.3pp for prior women's-merch buyers and +1.1pp for everyone else. Balance checks, Lin-adjusted effects, FDR-corrected interactions, and uplift models yield a policy that roughly doubles net value while sending 41% fewer emails.
"Wrapped, but honest" — a SQL-first DuckDB pipeline over Spotify's Extended Streaming History: 30-minute sessionization, skip-rate and taste-concentration (HHI) metrics, artist cohort retention, and a two-proportion z-test on whether shuffle is skippier. Built-in QA reconciles every raw row, and it runs end-to-end on a synthetic sample so no personal data is committed.
A beginner-focused Teamfight Tactics coach built on an original Monte Carlo expected-value engine. Describe a game state and the FastAPI solver simulates thousands of rollouts, then collapses the EV table into one move — roll, level, save, or stabilize — with a plain-English why. Includes a playable Arena that grades every decision post-game.
An end-to-end R report flagging at-risk customers across 10,000 records and 14 features. Logistic regression vs. Random Forest under 5-fold cross-validation, handling a 20.4% churn class imbalance, with decision thresholds tied to retention strategy.
Selected for Purdue's Market Readiness Incubator. As project manager, I designed the analytics-ready PostgreSQL (Supabase) backend — a relational schema with SQL CRUD plus multi-field search, sorting, and pagination, backed by integrity checks.
NASA Open Science Data Repository · Remote
Volunteer member of two of NASA OSDR's open-science Analysis Working Groups: AI/ML, which builds machine learning tools for spaceflight biology, and Brain. In Brain's Behavior & Neuroinflammation subgroup I work on linking a rodent behavioral phenotype from Rodent Research-1 to inflammasome-related neuroinflammation under spaceflight and ground-based radiation — pairing OSDR omics and behavioral datasets with environmental telemetry and RadLab radiation records. The subgroup is defining its research question and analysis plan.
Spiritual Data · Remote
Built a benchmarking harness comparing a proprietary model against baselines on 3+ metrics, designed evaluation splits and error slices to surface failure modes, and produced a stakeholder evaluation report aligned to business and compliance needs.
UC Berkeley · Lawrence Hall of Science · Berkeley, CA
Cleaned and prepared survey data from 700+ participants into analysis-ready datasets, then ran quantitative analysis (EDA, regression) on how social media use relates to public attitudes.
B.S. Informatics & Psychology — Double Major
A double major bridging data and human behavior — informatics for the technical foundation (programming, databases, analysis) and psychology for the research methods and statistics behind asking sharper questions of the numbers.
Open to full-time data analyst roles and collaborations.
yougi@yougijain.com