Case Study - AI engineering consulting for an education publisher

The problem

Education publishing has a lot of high-quality content sitting in formats that pre-date the LLM era. Textbooks, problem sets, instructor guides, the long tail of supplementary materials. The opportunity is to make that content useful in new ways: as the source of truth for tutoring assistants, as the corpus behind retrieval-augmented learning tools, as the raw material for adaptive content. The risk, from a publisher's seat, is putting any of that into systems that hallucinate, drift, or fail to cite back to the source. The trust bar is high for good reasons.

What we shipped

Senior consulting hours on the parts of the problem where I could help. The pattern I bring into engagements like this is the same one I bring to every AI project. Decide where the model gets to be creative and where it has to call a tool. Pin the model to verifiable sources. Build the eval rubric before you build the third feature. Treat the integration points as the actual product, not the demo.

Specifics on scope, deliverables, and outcomes are something I'm happy to walk through under NDA. I'd rather keep a careful page than a flattering one.

Outcome

The engagement was a successful one from a consulting standpoint. The team got senior AI engineering input at a point in their roadmap where the stakes were "make a decision we won't regret in 18 months," and the decisions held up.

If you're at an education company or a publisher and you're trying to figure out where LLMs fit into your content stack, that's a conversation I'm happy to have. I've now done some version of it more than once.