Multimodal E-commerce Retrieval System

A multimodal product-retrieval application supporting text and image queries.

  • Duration: July 2024 – November 2024
  • Technologies: Python, React, PyTorch, NLTK, RAG

View code on GitHub

Highlights

  • Built a web application for efficient e-commerce product retrieval using text and image queries.
  • Implemented multimodal feature fusion with Qwen-7B, hierarchical filtering with Sentence-BERT embeddings, and the Claude API for preprocessing.
  • Evaluated retrieval performance using BM25, TF-IDF, and BERT, achieving a score of up to 4/5 from a multimodal LLM judge.