Working on data systems. Mostly the parts that are too slow.
Noah Ball · data engineer / data scientist · NYC
Data infrastructure and AI systems at work. The rest of this I built to see if it would work.
work index
- 001 VantaPool Equity-swap analytics over about a billion rows of public disclosure data. ~190×
- 002 BallGPT Picks March Madness games. Backtested on 18 tournaments, 2008 to 2025. ~71%
- 003 yt-recap Ten years of my YouTube watch history, pulled apart in DuckDB. 30K+ videos
- 004 vivaria I told Claude to build whatever it wanted. --dangerously-skip-permissions
About
I think a lot of data problems are really patience problems. Somebody accepts a slow pipeline once, everyone builds on top of it, and five years later that wait is just how things are. An MS in Data Science taught me the models. Production taught me the model is rarely the hard part. It's upstream, in how the data is shaped and how fast it can move. Finding those spots and pulling them apart is the part of this I actually enjoy, on the job and on everything listed above.
toolbox
- Python
- DuckDB
- ClickHouse
- Postgres
- React
- Astro