Design a Visual Search Backend for a Boutique Luxury Marketplace
Overview
What this challenge is about.
You receive a catalog of 80,000 luxury items (image + sparse metadata) and a labeled query set of 300 user photos with hand-picked target items. Choose an embedding strategy (CLIP, DINOv2, or a fine-tuned vision backbone) and build a vector index. Evaluate recall@12 and qualitative match quality. The backend must return results within 200 ms p95. Deliverable is the trained or selected embedding pipeline, the retrieval API, and a merchandising-team brief on quality limits.
The Brief
What you'll do, and what you'll demonstrate.
Ship a visual-search backend with recall@12 above 0.7 on the labeled query set and p95 latency under 200 ms.
Earning criteria — what you'll demonstrate
- Select and evaluate vision embedding models for retrieval
- Operate a vector index at catalog scale
- Measure retrieval quality with recall@k and qualitative review
- Communicate retrieval-quality limits to a non-technical merchandising audience
Program Fit
Where this fits in your program.
Sharpens the same skills your degree expects you to demonstrate.
Skills
Skills you'll demonstrate.
Each one shows up on your verified credential.
Careers
Roles this prepares you for.
Real titles. Real skill bridges. Pick the one closest to your trajectory.
Machine Learning Engineer
Owning a visual-search backend end-to-end is a canonical MLE deliverable at marketplaces and e-commerce companies.
This challenge sharpens
- visual-search
- embeddings
- vector-search
Computer Vision Engineer
Tuning vision embeddings for catalog retrieval is increasingly a CV-engineer specialization at marketplaces.
This challenge sharpens
- embeddings
- clip
- retrieval-evaluation
AI Engineer
Standing up the retrieval API plus the metadata pre-filter and the merchandising brief is exactly the bundle AI engineers ship at small product teams.
This challenge sharpens
- visual-search
- clip
- retrieval-evaluation