← All work
AI & Data Solutions
Jewelry Recommendation System
A visual + metadata recommendation engine trained on 16,000+ products, retrained and redeployed automatically with zero manual steps.
- Role
- AI Engineer — modelling, pipeline, deployment
- Context
- E-commerce · Ashi Diamonds
- Category
- AI & Data Solutions
- Live
- Visit site ↗
Overview
A deep-learning recommendation engine for a large jewelry catalog that blends image similarity with structured metadata — style, type, colour, karat — to surface genuinely relevant products and lift engagement.
What I did
- Built a visual + metadata-based recommendation engine for jewelry products using deep learning.
- Trained the model on 16,000+ products, each with multiple images and rich metadata including style, type, color and karat.
- Used AWS SageMaker for model training, feature extraction, and FastAPI for endpoint deployment.
- Enabled automatic retraining and deployment via periodic jobs for keeping recommendations up to date.
- Optimized performance and scalability using FastAPI, S3, and Dockerized workflows.
- Significantly improved customer engagement by offering intelligent, relevant suggestions.
- Developed an end-to-end automation pipeline with zero manual interface.
Tech
Deep LearningAWS SageMakerAWS S3FastAPIDocker
Have a project like this in mind?
Start a conversation