EngineerPrep

Learn AI engineering through focused lessons, production incidents, hands-on projects, applied assessments, and a mentor that answers while you read.

Learn by reading what broke

Most AI engineering material explains how a technique works and stops there. The harder skill is diagnosing a system that is already running and behaving strangely, where nothing has errored and the answer looks reasonable. EngineerPrep teaches that skill directly: you are given a real incident, the evidence an on-call engineer would actually have, and the job of working out what happened before anyone tells you.

Production Failure Labs

Each lab is one incident, captured. An agent charged a customer twice and reported success. A support agent answered correctly for three weeks without ever calling the tool it was contracted to use. A trace showed a required lookup was skipped, and the lookup had run. You read the execution tree, the logs, the system state and the code, commit to a diagnosis, and then judge the remediation an agent proposes for it. Some of those proposals are wrong, and recognising which is the point.

The curriculum

Nine modules covering the things production LLM systems actually fail on: agents and tool calling, retrieval, memory, evaluation, prompting, structured output, security, and running any of it at a cost you can defend. Lessons are short and built around a worked example rather than a summary, and each one ends somewhere you can check your understanding.

Projects you can run

Downloadable Spring Boot projects that compile, start, and do the thing the lesson described. They are the finished state of a lesson brief rather than a scaffold, so you can read working code, change it, and watch the tests tell you whether you were right.

Research

Write-ups of failures worth understanding in detail, with the evidence that established them.

Membership

The incident briefs are open to read. The evidence, the diagnosis work and the labs themselves are part of a membership.