Straight out of my master's, I joined Atlantic Packaging's development team. I work on AI tools and on the integrations that connect the company's business systems, so data and work move between them without anyone copying it by hand.
Four months in, it's where I'm learning what production software actually takes: real users, real data, and systems that can't go down.
Five Value-at-Risk models forecast out-of-sample every trading day for ten years on a stock/bond/gold portfolio, then judged with the same backtests banks use to validate their own risk engines.
An Elo model played the 2024 draw 100,000 times. Its #2 pick, Sinner, won it.
Code →Sixty years of NOAA data, forecast in R. Modelling the series together cut sunspot error by 45%.
Code →A daily Databricks pipeline where Claude scores every Wilmington rental against what my roommates and I wanted.
Code →I'm a recent grad with a master's in Computer Science & Information Systems from UNC Wilmington, now a developer at Atlantic Packaging.
The thing I keep coming back to is the tech side of finance: quant research, systematic trading, and the models underneath them. When something catches my interest, whether that's markets, tennis, space weather or an AI pipeline, I build it and check it against real data.
Recruiters, fellow builders, anyone into quant or AI: I'm always happy to talk projects, ideas or opportunities.
lukas@lukas-nilsson.com