OpenHack: Building an AI Legal Assistant with Azure OpenAI & AI Search
Build a production-ready AI legal research assistant using Azure OpenAI and AI Search in this OpenHack challenge. Design RAG architecture, implement document retrieval, prevent hallucinations, and secure for enterprise use.
View badge details
About This Course
Take on the role of an engineering team hired by Meridian & Associates LLP, a mid-size law firm with over 200 attorneys, to build "MeridianAI" an internal AI-powered legal research assistant using Azure OpenAI and Azure AI Search. This OpenHack-style challenge course puts you in a realistic scenario where you must design, build, and harden a production-ready RAG (Retrieval-Augmented Generation) system.The firm's attorneys spend an average of 3 hours per day searching through case summaries, contract templates, regulatory guidance memos, and client engagement letters. Your mission is to build an AI assistant that answers attorney questions grounded exclusively in the firm's document library where hallucinations are unacceptable, every answer must be traceable to a source document, and the system must comply with strict data security policies for sensitive client information.This is not a tutorial. You will make real architectural decisions, provision Azure infrastructure, build a document ingestion and retrieval pipeline with Azure AI Search, integrate Azure OpenAI for intelligent responses, implement grounding and citation strategies to prevent hallucinations, and secure the system for production use. You'll debug real problems and deliver a working system that meets enterprise requirements.Designed for cloud engineers, AI engineers, and developers with basic Azure experience who want hands-on experience building RAG-based AI applications. This challenge-based format emphasizes problem-solving and engineering judgment over step-by-step instructions.
Course Curriculum
1 Lessons
01
Challenge
OpenHack: Building an AI Legal Assistant with Azure OpenAI & AI Search
Meridian & Associates LLP needs an AI-powered legal research assistant built on Azure OpenAI and Azure AI Search. Teams will provision infrastructure, build a RAG pipeline, add citation tracking and content safety, and harden the system for production use — all through progressive engineering challenges.
Challenge 1 — Foundation: Provision the AI Platform
Deploy the core Azure services needed for the AI legal assistant: Azure OpenAI (with GPT-4o and text-embedding-ada-002 models), Azure AI Search, and Azure Blob Storage. Verify all services are operational and properly configured.
~45 min
Challenge 2 — Ingestion: Feed the Knowledge Base
Upload Meridian's document library to Blob Storage, configure an AI Search indexer with a skillset for document cracking and chunking, generate embeddings, and populate a vector search index. The index must support both keyword and vector search.
~60 min
Challenge 3 — Grounded Answers: Build the RAG Pipeline
Build the retrieval-augmented generation pipeline that queries the AI Search index, retrieves relevant document chunks, and sends them as context to Azure OpenAI to generate grounded answers. Implement a system prompt that constrains the model to only answer from provided context.
~75 min
Challenge 4 — Trust but Verify: Citations & Guardrails
Add citation tracking so every AI response references the source document and section. Integrate Azure AI Content Safety to filter inappropriate inputs and outputs. Implement grounding detection to identify and flag responses that may not be fully supported by the source documents.
~75 min
Challenge 5 — Production Ready: Secure, Monitor, Scale
Harden the system for production: switch to managed identity authentication (eliminate API keys), enable Application Insights for monitoring and tracing, configure diagnostic logging, set up rate limiting, and validate the system handles concurrent attorney queries under load.
~75 min