Data Science + Statistics at UC Berkeley. I build pipelines that make noisy data usable.
Pipelines and tools I’ve built on my own time. Each one has a write-up on how it works.
An LLM assistant that lets anyone ask questions about market data in Slack
A cron-scheduled ETL pipeline lands 1-minute equity bars in date-partitioned Parquet, and an MCP server lets a Slack chatbot query it in plain English.
Live ADS-B telemetry turned into clean, analysis-ready flight features
An end-to-end pipeline that pulls a live global ADS-B feed, validates every record with Pydantic, derives flight-activity features in Polars, and warehouses the results in DuckDB.