Digital Asset Management Platform

Local Shot, a media production company, had terabytes of video footage and assets scattered across Dropbox accounts, external hard drives, and old archive disks. When a client requested footage from a 2019 shoot, finding it could take hours.

The 5-person team had no consistent way to search content. File names were cryptic. Metadata was inconsistent. Some drives weren't even plugged in. Older footage on cold storage was essentially lost since nobody remembered what was on each archive.

They needed a single source of truth where anyone on the team could search 'downtown LA night shots' and actually find relevant footage, regardless of where it was originally stored.

I built a centralized asset management system on S3 with a React frontend. Bulk uploads go through an SQS queue that processes files asynchronously, generating thumbnails with FFmpeg for video files and extracting technical metadata.

The key feature is AI-powered content tagging. Each uploaded asset runs through a vision API that detects scene content, objects, text, and generates searchable descriptions. A video of a beach sunset gets tagged automatically with 'beach, ocean, sunset, golden hour, waves' without manual input.

For their archived footage on cold storage, we implemented a tiered system: metadata and thumbnails are always available for search, but the full file can be on Glacier until someone actually needs it. This kept storage costs manageable for terabytes of archives.