Computer Vision / AI · Prototype
Visual Tagger
Hybrid ViT + CLIP approach for structured image insights with FastAPI backend.
Role: Individual project
Problem
Extract structured, actionable insights from images with hybrid vision models.
Approach
Async FastAPI backend combining ViT and CLIP for classification and zero-shot tagging.
Engineering decisions
- • Hybrid approach for better coverage
- • API-first design for reuse
Challenges & trade-offs
- • Model selection and inference performance
Notes
- • Prototype stage.