AWS Generative AI Portfolio

This repository contains clear, production-oriented implementations of Generative AI serverless architectures on AWS. It demonstrates the practical application of Large Language Models (LLMs), Vector Databases, and Agentic workflows to solve real-world problems.

Key Projects

1. Serverless GenAI QA App

A full-stack RAG application that allows users to query a knowledge base using natural language.

• Tech Stack: Amazon Bedrock (Agents), Amazon Aurora PostgreSQL (pgvector), AWS Lambda, React, AWS Amplify.
• Key Features: Retrieval-Augmented Generation, vector similarity search (HNSW), and an interactive chat interface.

2. Bedrock RAG System

A deeper dive into the backend implementation of Retrieval-Augmented Generation.

• Tech Stack: Amazon Bedrock (Titan models), SQL-based Vector Store.
• Focus: Configuring high-performance vector indexes (GIN/HNSW) and secure knowledge base integration.

3. Conversational RAG Agent

An advanced agentic workflow capable of multi-turn conversations.

• Tech Stack: Amazon Bedrock Agents, Aurora PostgreSQL (1024-dim Layout).
• Key Features: Context retention, orchestrated retrieval, and higher-dimensional embedding support.

4. GenAI App with Terraform

A serverless architecture for document ingestion and QA, provisioned entirely with Terraform.

• Tech Stack: Terraform, Amazon Bedrock (Knowledge Bases, Guardrails), AWS Lambda, Amazon API Gateway.
• Key Features: Infrastructure as Code, automated ingestion pipeline, and responsible AI guardrails.

Technologies Used

• Generative AI: Amazon Bedrock, Amazon SageMaker, Titan Embeddings, Foundational Models (FM).
• Database: Amazon Aurora PostgreSQL (pgvector extension).
• Compute: AWS Lambda (Serverless Python).
• Infrastructure: AWS IAM, Secrets Manager, S3.
• Frontend: React.js, AWS Amplify.

---

Built to demonstrate scalable and secure AI solutions compliant with industry standards.