AWS
Bedrock LLMs
RAG
Retrieval Pipeline
pgvector
Vector Store
E2E
GenAI Platform
Enterprise-grade GenAI data engineering platform using AWS Bedrock for LLM inference, pgvector for semantic search, and a full RAG pipeline over a cocktail knowledge base. Demonstrates the full AWS GenAI stack from ingestion to retrieval to generation.
The full AWS Bedrock RAG stack — embeddings, vector store, retrieval, generation — in one production-ready platform.
Tech Stack
PythonAWS BedrockpgvectorRAGPostgreSQLLangChainboto3DockerFastAPI
Features
pgvector Store
PostgreSQL + pgvector for production semantic search. No managed vector DB required.
Bedrock LLM
Claude and Titan models via AWS Bedrock API. Swap models without changing application code.
RAG Pipeline
Full retrieval-augmented generation: chunk → embed → store → retrieve → generate.
Enterprise Stack
Production patterns: retry logic, observability hooks, cost tracking per query.
How It Works
01
Ingest
Cocktail dataset → chunked documents → Bedrock embeddings
→
02
Store
Vectors → pgvector PostgreSQL with cosine similarity index
→
03
Retrieve
User query → embed → top-k semantic search
→
04
Generate
Retrieved context + query → Bedrock Claude → grounded response