Join us to explore how to transform a traditional financial application into an intelligent, AI-powered solution using PostgreSQL and Azure. In this hands-on session, you'll learn how to integrate generative AI to deliver high-quality responses to financial queries through a smart copilot that leverages PDF-based Statements of Work (SOWs) and invoices.
We'll guide you through:
- Performing AI-driven data validation
- Using the Azure AI extension for PostgreSQL
- Implementing vector search with DiskANN indexes
- Enhancing results with semantic re-ranking
- Building with the LangChain agent/tools framework
- Applying GraphRAG on Azure Database for PostgreSQL
By the end of the session, you'll have built a practical, end-to-end AI-driven application---and gained the skills to apply these techniques to your own data and projects.
📌 This session is a part of a series. Learn more here: https://aka.ms/BuildAIapps/PostgreSQL
#MicrosoftReactor #learnconnectbuild
Chapters:
00:00 - Welcome & Housekeeping
00:57 - Meet the Speakers: Aaron Wislang & Ismael Mejia
01:48 - Series Recap & Today’s Use Case
03:20 - Automating Contract & Invoice Validation
05:02 - Architecture Overview
08:08 - Document Flow & AI Services
10:03 - Document Examples: SOWs & Invoices
13:23 - Setting Up the Lab Environment
16:32 - Exploring the PostgreSQL Schema
20:05 - Database Setup & Extensions
25:00 - AI-Driven Document Validation
33:02 - Embeddings, Indexing & Vector Search
43:01 - Chunking & Summarization with Azure AI
47:00 - Prompt Engineering for Validation
54:03 - Building the AI Copilot Chat
1:00:00 - LangChain Integration & Agent Tools
1:04:00 - Live Demo: Chat UI in Action
1:07:16 - Improving Accuracy with Semantic Ranking
1:09:01 - GraphRAG & Graph Queries in PostgreSQL
1:12:21 - Final Thoughts & Resources
[eventID:26070]