Bayesian Inference: Priors, MCMC, and Deciding When to Dig

Deep End Era. Engineering with Sankalp · Oct 28, 2026 · 1h 3m

Three inline-inspection readings say a corroding pipe has seven years left. They're probably right, but "probably" isn't good enough near a river crossing. This episode follows Ulla Brennan and Desmond Achterberg (both fictional) as they turn three noisy numbers into a probability, then a decision. Along the way it covers why random walks stumble in high dimensions, how a skateboard in a bowl became the standard sampler, why divergences matter, how partial pooling borrows strength across a pipeline, and how treating failed searches as evidence found a wreck 14,000 feet down. Cross-references…