* Semantic Search and RAG with Vector Databases and LLMs

Summary

As someone passionate about fitness, I realized that over time I had curated a significant amount of fitness content from my favorite scientists, coaches, content creators, etc. I wanted to create a RAG pipeline to query trusted fitness content that I already saved, rather than searching the web and needing to sift through the noise.

Brief details

• Implemented vector-based similarity search using Weaviate, integrating a Dockerized vector database and text2vec transformer module to efficiently query vectorized JSON data for semantic search capabilities
• Utilized Hugging Face and Lightning AI to design and run a Retrieval-Augmented-Generation pipeline on GPU-accelerated cloud platforms, enabling natural language queries on customized datasets and PDFs