How charuka-g/recommendation-system-for-ecommerce-aws works
This repository implements a two-stage recommendation product: training and preprocessing notebooks prepare retrieval candidates and a ranking model, while a FastAPI endpoint retrieves a user's candidates, enriches them with features, ranks them, and returns the top results. The graph includes the README-documented training, registry, and Feature Store capabilities as well as the sampled API call flow. Notebook internals were not sampled, so their relationships are based on README statements.
Main entry points: inference_api.py Inference API · 01-Train-FM-Model.ipynb Candidate Generation · 02-Train-XGBoost-Model.ipynb XGBoost Training · 00-Setup-Configuration.ipynb AWS Setup
Components 16
Model Training
- AWS Setup
00-Setup-Configuration.ipynb - MF Training
01-Train-FM-Model.ipynb - Candidate Generation
01-Train-FM-Model.ipynb - XGBoost Training
02-Train-XGBoost-Model.ipynb
Feature Preparation
- Feature Processing
03-Glue-Feature-Store-Processing.ipynb
Recommendation Serving
- Inference API
inference_api.py - Candidate Lookup
inference_api.py - Feature Enrichment
inference_api.py - XGBoost Ranking
inference_api.py - Top-K Results
inference_api.py
AWS Services
- S3 Artifacts
- SageMaker Registry
- DynamoDB Candidates
- Feature Store
- SageMaker Endpoint
Other components
- User
Connections 18
- User → Inference API: requests recommendations
README.md - AWS Setup → MF Training: configures
README.md - AWS Setup → XGBoost Training: configures
README.md - MF Training → Candidate Generation: extracts embeddings
README.md - Candidate Generation → S3 Artifacts: saves artifacts
README.md - Candidate Generation → SageMaker Registry: registers model
README.md - Candidate Generation → DynamoDB Candidates: stores candidates
README.md - XGBoost Training → Feature Store: joins features
README.md - XGBoost Training → SageMaker Endpoint: deploys model
README.md - Feature Processing → Feature Store: processes feature data
README.md - Inference API → Candidate Lookup: calls
inference_api.py - Candidate Lookup → DynamoDB Candidates: reads candidates
inference_api.py - Inference API → Feature Enrichment: calls
inference_api.py - Feature Enrichment → Feature Store: fetches records
inference_api.py - Inference API → XGBoost Ranking: calls
inference_api.py - XGBoost Ranking → SageMaker Endpoint: invokes
inference_api.py - Inference API → Top-K Results: sorts and limits
inference_api.py - Top-K Results → User: returns recommendations
README.md