Agentic AI Foundations for Product Builders
Build a dinner-assistant prototype. Learn to structure requests, use evidence, connect tools, evaluate behaviour and require approval.
Six modules, 24 lessons. Plan 4–6 hours per module. Progress saves in this browser. Labs use fictional sample data and simulated orders.
AI Foundations: Understand and Structure User Requests
Connect generative AI foundations to product decisions: prompts, context, structured responses, and a tested request-understanding workflow. Four lessons plus a guided build; allow 4–6 hours.
Retrieval: Ground Recommendations in Reliable Information
Retrieve useful menu evidence, preserve its context, filter real constraints, and explain recommendations with sources.
Agent Workflows: Connect Tools and Manage State
Define tool contracts, inspect state transitions, recover from failures, and understand when frameworks or multiple agents help.
Model Adaptation: Choose Prompts, Retrieval, or Fine-Tuning
Diagnose failures before choosing an intervention, understand LoRA and QLoRA, and compare local-model tradeoffs.
Evaluation: Measure Quality and Diagnose Failures
Build a small test set, use appropriate graders, inspect traces, and compare changes without hiding critical failures.
Safety and Release: Protect Users and Require Approval
Keep untrusted content separate from authority, minimize sensitive data, enforce approval, and demonstrate the complete prototype honestly.