This program equips data analysts and technical professionals to design AI solutions that stand up to real business scrutiny—measured, explainable, compliant, and optimized for impact. Across 11 hands-on short courses, you’ll build and evaluate conversational AI (including retrieval-augmented generation), explain black-box models for executive audiences, and move from descriptive analytics to prescriptive decision intelligence. You’ll also learn to diagnose operational problems with root-cause methods, apply modern optimization approaches (linear programming, mixed-integer methods, genetic algorithms, and reinforcement learning), and deploy real-time decision platforms that meet tight SLAs. The program rounds out with causal inference techniques to estimate true business impact, plus ethical AI, debiasing, privacy, and compliance practices to reduce risk and increase trust. Each course emphasizes practical deliverables, clear metrics (quality, fidelity, fairness, latency, robustness), and stakeholder-ready communication—so your work translates into measurable outcomes, not just model performance.
Applied Learning Project
You’ll complete project-based work that mirrors real analytics and AI delivery: build a RAG chatbot prototype and assess usability with a SUS survey; improve dialog performance using flow metrics like fallback rate and turn length; extract entities from support tickets and report precision/recall; and compare TF-IDF vs. embeddings for classification. You’ll also explain black-box models using SHAP, LIME, counterfactuals, and surrogates—then quantify fidelity and stability. Optimization projects include linear programming for product mix, mixed-integer logistics routing, elasticity-based pricing simulation with guardrails, and heuristic vs. RL/GA approaches for inventory and supply chain. You’ll implement causal methods (PSM, assumption checks, PC algorithms, bootstrap stability), fairness audits and mitigation, differential privacy with privacy budgets, and a streaming decision pipeline (Kafka/Spark) validated under load.
