This is the flagship: twenty-four hours across one hundred and eighty lessons, taking a developer who has never called a model API to someone who can design, build, evaluate and operate an AI product end to end. It is structured as eight parts, each with its own project, culminating in a single deployed application you build across the whole program.
Part one covers the API surface and cost fundamentals. Part two is prompting as engineering, with versioned templates, structured output and composition. Part three is retrieval, from embeddings through chunking to hybrid search and reranking on an awkward corpus. Part four is tool use and bounded agent loops. Part five is evaluation, the part most self-taught engineers skip and later regret.
Part six is architecture and operations: queues, streaming, fallbacks, caching, routing and per-request budgets. Part seven covers safety, injection defence, refusal design and the audit trail you need when a customer complains about an answer. Part eight is delivery, taking the application through deployment, monitoring, a staged rollout and a genuine post-launch review of production transcripts.
The program is opinionated: it teaches one good path rather than surveying every option. What you finish with is not a certificate but a running system, a test suite, an eval harness and a cost model you can defend in a meeting.