Production Decision Models: From Prediction to Action

What you'll learn

• Frame any business decision as actions, objectives, constraints, and uncertainty before choosing a technique
• Calibrate model probabilities and convert them into cost-sensitive, expected-value decisions
• Quantify uncertainty with conformal prediction and quantile models that come with coverage guarantees
• Formulate and solve linear and mixed-integer programs, and avoid the traps of naive predict-then-optimize
• Estimate causal effects and uplift so you target the customers an action actually changes
• Run bandits for explore-exploit decisions and evaluate new policies offline from logged data
• Combine decision tables, rules engines, and LLM components with guardrails and human review
• Measure decision quality, run policy experiments, and monitor for drift and feedback loops
• Ship decision services with fairness checks, explanations, audit trails, and simulation-based testing