Blog
Posts on ML, math, and the art of data alchemy.
-
From five million variants to ten candidates
Part two of a series on the 2026 MVA hackathon: what the prioritization pipeline does to the VCF, one invented row followed through every gate, and why the filters set the ceiling.
-
A rare disease is a ranking problem
Part one of a series on the 2026 MVA hackathon: what the disease is, the biology you need to read a VCF, and what the task asks for.
-
The crisis of confidence in statistical purity
Why state-of-the-art models are confidently wrong, and what calibration and conformal prediction do about it.
-
When data doesn't play fair
When 99.9% accuracy means nothing: fraud, rare diseases, and learning from lopsided data.
-
Is statistical learning 'all we need'?
A deep dive into the foundations of machine learning and the enigma of deep learning generalization.