SlackShots
A Slack-integrated media platform that automates hundreds of weekly uploads and adds MobileCLIP-powered semantic retrieval through MongoDB Atlas Vector Search.
Role
End-to-end (workflow design, full-stack development, Slack integration, and semantic retrieval)
Timeline
April 2024 - Present
Problem
Uploading 200 to 300 images into Slack each week was slow, fragile, and tedious. Native uploads had to be done in small batches, took too long, and frequently crashed during the process.
Approach
Built a Slack-connected platform that batches uploads, lets users choose the destination workspace and channel, and centralizes previously uploaded media. Extended it with locally generated MobileCLIP embeddings, MongoDB Atlas Vector Search, resumable background indexing, and ranked retrieval over Slack-hosted source images.
Results
- Reduced recurring upload and media-management work by approximately 30%.
- Supports hundreds of images each week through a single streamlined workflow.
- Added semantic retrieval without introducing a separate source-image storage layer.
Impact
- Built around a recurring weekly workflow with hundreds of images per session.
- Uses MongoDB Atlas Vector Search and resumable indexing rather than a mocked search experience.
- Keeps Slack as the source of truth while providing a faster interface for discovery and review.
Gallery
