Pair-programming with a coding agent inside a Marimo notebook to build a Bayesian hierarchical model of protein melting points, live.
Live demo recorded at Data Driven Pharma East 2026 in Boston. I pair-programmed with a coding agent (Cursor + Composer 2.5) inside a Marimo notebook to build a Bayesian hierarchical model of protein melting points in real time. We load and explore protein meltome data, build a four-parameter logistic (4PL) curve with PyMC, estimate melting temperatures with full uncertainty quantification, expand to a hierarchical model across yeast proteins, and rank proteins by thermal stability. Along the way I share lessons on when to go fast versus slow down with coding agents, sequential versus parallel exploration, data provenance and reproducibility, and model cost considerations.