Cloud-Native RAG System with LLM Integration

This personal initiative demonstrates the end-to-end development and deployment of a sophisticated Retrieval Augmented Generation (RAG) system. The project focuses on building a scalable, cloud-native solution capable of performing intelligent question-answering over custom datasets by integrating Large Language Models with external knowledge bases. The entire infrastructure is meticulously defined and managed using Terraform on AWS.

The core RAG pipeline, orchestrated with LangChain, ingests documents, creates embeddings (via LLMs), stores them in PostgreSQL with the pgvector extension, and enables semantic retrieval to provide contextually relevant information to LLMs (from OpenAI and Anthropic) for generating accurate and nuanced answers. The backend services, combining Python for AI logic and TypeScript/Node.js for API layers, are deployed as containerized applications on AWS Fargate and serverless functions on AWS Lambda. This architecture includes a full suite of AWS services like API Gateway, Application Load Balancers, S3, IAM, VPC, and CloudWatch.

A dynamic Next.js frontend allows users to interact with the RAG system, submitting queries and receiving AI-generated responses. The project features a robust CI/CD pipeline using GitHub Actions for automated builds, tests, and deployments. Documentation is maintained with Docusaurus 2.