Assistants, GPT apps and workflow automation
Most ChatGPT projects die between the demo and the deadline. This track is about the unglamorous middle: contracts the model can't break, retrieval that finds the right paragraph, and an evaluation harness that tells you when a prompt change made things worse.
Ordered the way we teach it. Private workshops reorder and reweight this against what your team already knows.
Moving from a clever prompt to a specified one — inputs, refusals, edge cases and the failure modes you only see at volume.
Treating the model as a component with a typed interface, so downstream code can parse it without a pile of defensive branches.
Chunking, embeddings and re-ranking — and how to tell whether retrieval or generation is the thing that's actually failing.
Packaging a working system so operations, sales or support can use it daily without a developer in the loop.
The three numbers that decide whether a feature ships. Where to cache, what to stream, and which model tier the task honestly needs.
Building the graded test set that turns 'it feels better' into a number you can put in a pull request.
Teams often pair a model track with Agentic Systems — the tool design and evaluation material applies whichever provider you settle on.
Send over your stack, your group size and roughly when you'd like it. We'll come back with a curriculum outline and a date.