Stage 2 · Lesson 7 of 7
Lineage: from prediction back to data
is the recorded chain from a single prediction back to everything that produced it: the that answered, the training run that created it, the Git commit, the and the data version. For an LLM app it also covers the and the index version.
You get lineage by logging the model version with every prediction and by having your runs record commits and data hashes, which DVC and make easy. When a user reports a bad result, lineage turns "what changed?" into a lookup rather than an investigation. Regulated industries such as banking and health increasingly ask for it.
Common traplogging predictions without the model version, which breaks the chain at the first link.
A question to check yourself, an AI practice session and 1 cheat sheet entry come with the full roadmap. Unlock
Resources for this stage
Start with
- CourseFreeMade With ML
Goku Mohandas · Advanced
Free: the versioning and data testing lessons, built around one real project