What Makes an AI App Different from a Script
By the end of this module you can name what makes an AI application harder than a script: an output you cannot fully predict, a cost per call, and a failure mode that returns confident nonsense.
Generative AI Application Development with Python
A small AI assistant app with structured input, structured output, one safe tool, mock mode, tests, and a release checklist.
API Keys, Environment Variables, Costs, and Safe Local Setup
By the end of this module you can set up keys and environment variables safely, estimate what a feature will cost per call, and avoid the mistake that puts a secret in version control.
Messages, Instructions, and Response Handling
By the end of this module you can structure messages and system instructions deliberately, and handle a response as data to be checked rather than text to be trusted.
Prompt Templates and Input Validation
By the end of this module you can build prompt templates with validated inputs, so user text cannot silently rewrite your instructions.
Structured Outputs with Schemas
By the end of this module you can request structured output against a schema, and handle the case where the model returns something that does not match it.
Tool and Function Calling with External Actions
By the end of this module you can let a model call your functions, and design each tool so that a wrong call is cheap, reversible, and visible.
Conversation State and Memory Boundaries
By the end of this module you can manage conversation state deliberately, deciding what to carry forward and what to drop before the context window decides for you.
File Input, Summarisation, and Extraction Workflows
By the end of this module you can build summarisation and extraction workflows over real files, and check the output against the source rather than against how good it sounds.
Error Handling, Retries, Rate Limits, and Fallback Messages
By the end of this module you can handle rate limits, timeouts and failures so that a degraded provider produces a useful message rather than a stack trace.
Simple App UI with FastAPI or Streamlit
By the end of this module you can put a usable interface on an AI feature with FastAPI or Streamlit, and choose between them on what the application needs.
Testing AI App Behavior with Deterministic Fixtures
By the end of this module you can test AI behaviour with deterministic fixtures, so your suite is fast, free, and fails for a real reason.
Privacy, Logging, and Safe Deployment Boundaries
By the end of this module you can decide what must never be logged, what must never leave your infrastructure, and how to deploy an AI feature without leaking either.
Capstone: Useful AI Assistant with Evaluation Checklist
In this capstone you will ship a genuinely useful AI assistant with structured outputs, real error handling, a deterministic test suite, and an evaluation checklist someone else can run.
First-release boundary
This is a text-first course shell. Leonardo/image creation, image QC, dependency-heavy runtime QA, external review, and learner pilot evidence remain separate later quality steps.