Building AI Applications

Put a model behind a product: the request and its parameters, structured outputs and function calling, retrieval, and the cost, latency and privacy decisions that ship.
- Islands
- 4
- Lessons
- 25
- Exercises
- 143
- Island 1
Models and the API
The request an application actually sends: messages and roles, the parameters that change the answer, the token budget that sets the bill, and the errors a production client must survive.
- Island 2
Structured Outputs and Function Calling
Get a shape back instead of prose, run the tool loop from the application's side, handle results and errors without losing the thread, and validate everything before it touches your data.
- Island 3
Retrieval and Context
Give the model the right few pages instead of the whole library: when to retrieve, how chunks and embeddings work, how to assemble a context the model can cite, and how to tell when retrieval is the wrong tool.
- Island 4
Cost, Latency and Shipping
The decisions between a demo and a product: which model for which request, caching, streaming and latency, what may enter a prompt and a log, and the checks that run before and after launch.