← Selected work 
Own build
Footage library audit and AI tagging
A local tool that audits a 900-clip footage drive, finds duplicates and orphaned files, detects trips, and tags every clip with a vision model, without ever moving a file destructively.
The situation
Years of drone and camera footage on one drive: several devices, inconsistent naming, duplicates from re-imports, orphaned metadata files, and no way to find the good clips.
What we built
- An audit pass that inventories devices, dates, duplicates, and orphan files
- Trip detection that groups clips into shoots automatically
- Vision-model tagging of every clip, with a ranked gallery for browsing
- A virtual organization layer with an undo trail, and a hard rule that files on the drive are never moved or deleted by the tool
What carries over
The same shape as a data pipeline for any messy asset library: inventory, dedupe, classify, and organize behind a safety rail.
Result
A searchable, ranked library built from a drive that used to be a folder of dates.