Self-initiated work sample: a source-pack CSV QA check using synthetic data.
I built a small JavaScript checker that flags duplicate organization names, missing required evidence fields, malformed URLs, and stale or future source-review dates. In the five-row fictional fixture, one row passes structure and four are held for correction. A structural pass never verifies a public claim, an official source, a buyer, or deliverability. Every real source still needs human review.
The input and results contain no client or private data. AI assisted code drafting; I ran the fixture and reviewed the output. This QA step supports my bounded source-backed research service: https://contra.com/s/QBg2evNW-public-source-prospect-research-pack