Watching lessons is only half of learning. These four projects turn the ideas from the course into working systems: a question-answering system over your own documents, an agent that uses tools, an assistant that remembers, and a set of evals that tell you whether your prompts actually work.
Each project comes with a video that walks through the build. Watch it, code along, then try the same thing again on your own documents, tasks or prompts. Each project also lists the modules it uses, so you can go back and revise anything that feels shaky.
Build retrieval-augmented generation in Python step by step: split and index your documents, retrieve the best passages for a question, and have the model answer from them. Then point it at documents of your own and check that the answers stay grounded. Uses modules 7, 8 and 9.
Video 2h 33minBuild a minimal agent in plain Python: a model call, a tool the model can request, a loop that feeds the results back, and a to-do list for planning. Then give it a tool of your own and watch what lands in its context. Uses modules 3, 8 and 11.
Video 19minCode an email assistant that remembers: store facts about the user (semantic memory), past examples (episodic memory) and its own instructions (procedural memory), and bring them back into the prompt when needed. Uses modules 4, 10 and 11.
Video 1h 31minTake a prompt you use, collect real outputs, write notes on where it fails, and turn the most common failures into pass/fail checks, including an LLM judge you test against your own judgement. Uses modules 2, 6 and 12.
Video 54min