Adapts existing models to a specific domain, and knows when not to.
Why this one is hard to judge
Knowing when fine-tuning is the wrong answer is worth more than knowing how to do it, and it is harder to demonstrate.
What to ask for
Ask for a model they fine-tuned and what specific problem the base model could not solve.
Ask how they built or chose the training data, and who checked its quality.
Ask for an example where fine-tuning made something worse, and how they noticed.
Ask how they decided fine-tuning was the right approach instead of prompting or retrieval.
The mistake most hirers make
Hirers assume fine-tuning is always the answer to a model behaving badly, when a better prompt or better retrieved context often fixes it more cheaply. A candidate who defaults to fine-tuning without trying simpler options first is solving the wrong problem. The harder skill is knowing when not to fine-tune at all.
What good looks like after 90 days
A clear, honest comparison of fine-tuning against the simpler options that were tried first. If fine-tuning was the right call, a model that measurably improved on the specific problem it targeted. A repeatable process for building and checking training data.
How we assess it
Against a rubric that is published in full, on evidence the practitioner supplies and a reviewer checks. Where something has not been verified, the profile says so.