From correlation-based periodontal risk prediction (and watching it break in practice) to Stanford Causal AI: DML, Proximal Inference, Dynamic Treatment Regimes—moving from “predicting risk” to “estimating what works, for whom, and when.” That work made the pattern obvious: healthcare and manufacturing share the same failure mode—the Prediction Trap (great forecasts, weak outcomes). So I pivoted to the Intelligent Shopfloor: causal + adaptive decision systems that turn signals into transparent, outcome-improving interventions.