RAG-based AI App using Azure AI Foundry

This project was developed as part of Module 5 of the Microsoft AI Skills Challenge: Create agentic AI solutions with Azure AI Foundry. It showcases a Retrieval-Augmented Generation (RAG) application that enables querying custom PDF data using OpenAI models deployed on Azure.

What I Learned

• How to upload and manage custom datasets in Azure AI Foundry
• How to configure and use Azure AI Search for information retrieval
• How to implement hybrid search (semantic + keyword) with vector indexing
• How to integrate OpenAI's embedding and chat models in a RAG pipeline
• How to use .env for securely managing credentials and configuration
• How Retrieval-Augmented Generation (RAG) improves factual accuracy in LLM outputs

Project Highlights

• Uploaded travel brochures (PDFs) as custom data into Azure AI Foundry
• Created a vector index using Azure AI Search, enabling semantic retrieval
• Configured hybrid search to combine vector similarity and keyword scoring
• Used text-embedding-ada-002 for generating embeddings
• Passed search results as context to a gpt-4o chat model for enriched answers
• Configurations were managed using a .env file and excluded via .gitignore
• Screenshots included to demonstrate outputs in Azure

Screenshots

• (assets/configuration-file.png)
• (assets/chat-demo.png)
• (assets/chat-code-powershell.png)

Reference

• Microsoft Learn: https://learn.microsoft.com/en-gb/collections/50wkaqtq50egz3?ref=collection&listId=60yka7t2o8od52&sharingId=6A9F03F25E12DA9E&wt.mc_id=aiskillsfest_msftlearn_training_wwl_challenge_tech