This role requires a combination that is genuinely rare: research-grade statistical depth, the software discipline to implement models that engineers can maintain, and the client communication skills to present methodology, defend assumptions under live challenge, and adjust the model narrative when the room requires it. Take models written as equations -- likelihood, priors, constraints, forward maps -- and implement correct, tested, reproducible estimators in Python using probabilistic programming frameworks; derive estimators from first principles and diagnose identifiability, partial identification, and convergence.