An evening reproducing an open-source decision model's published numbers on a laptop - which claims held, which did not, how Metal compared to a T4, and the one measurement that turned into an upstream pull request.
Can you exploit Canada's APPR flight delay compensation rules for profit? I built a data pipeline, trained two ML models, and ran the numbers on 650 flights.