Case study

RAG-QA on AWS

Retrieval-Augmented QA, fully CI/CD

70B

params (LLAMA 3.1)

CI/CD

ECR + App Runner

Docker

reproducible deploys

The problem

Stand up a retrieval-augmented QA service on a 70B model that's reproducible and deployable — not a notebook demo that dies when the kernel restarts.

Vector retrieval over a document corpus feeds AWS Bedrock LLAMA 3.1-70B, grounding answers in source context.

Dockerised and shipped to AWS ECR + App Runner through GitHub Actions — every push produces a reproducible deploy.

LangChainFAISSAWS BedrockLLAMA 3.1-70BDockerGitHub Actions