AI Engineering Portfolio Artifact

Publish a public-safe case study without confidential data, prompts, credentials, or exploitable control details.

Include the problem and non-AI baseline; architecture and rejected alternatives; dataset and provenance summary; evaluation design and representative failures; safety and threat analysis; latency and cost evidence; release, monitoring, fallback, and rollback; limitations; and the changes produced by review.

Link privately retained evidence by stable artifact IDs. The reviewer must be able to distinguish verified work from assisted work through the AI-use declaration and oral defense. Marketing claims without reproducible evidence fail.