Build a Deep Research Agent for Enterprise Data with Scalable Performance

Brought to you by Microsoft and NVIDIA

Move beyond foundational concepts and dive into production-grade deployment of enterprise research agents. This session explores advanced architecture patterns for scaling NVIDIA AI-Q Blueprints and Nemotron models on Azure Kubernetes Services (AKS), focusing on real-world optimization strategies, performance tuning, and operational best practices.

Learn how to leverage NVIDIA NIM microservices alongside Nemotron's open-source models to build intelligent agents that reason across complex enterprise data, with practical insights on cost optimization, reliability, and multi-agent orchestration in Kubernetes environments

📌 This episode is a part of a series. Learn more: https://aka.ms/AIAgentsApps/y-MSFT

Chapters:
1:43 Welcome & Housekeeping
1:58 Code of Conduct Reminder
2:10 Chat & Links Overview
2:15 Recording Information
2:23 Speaker Introductions
2:48 Session Agenda
3:41 NVIDIA & Microsoft Collaboration
4:50 NVIDIA Hardware on Azure
5:31 NVIDIA AI Enterprise Overview
6:16 NeMoTron Open Model Family
7:57 Model Efficiency & Openness
8:20 NeMoTron 3 Architecture
9:57 Reinforcement Learning & Context Window
10:50 NIM Containers Overview
12:17 Open‑Source & Domain‑Specific NIMs
12:26 NVIDIA Enterprise on Azure Marketplace
17:27 Blueprint Overview
18:06 Blueprint Components
19:46 Deep Researcher Workflow
21:19 Architecture Walkthrough
23:05 Deep Researcher UI Demo
25:07 AKS Deployment Intro
26:05 RAG & AIQ Blueprints in AKS
28:00 Observability & Metrics Overview
29:04 Kubernetes & Telemetry
30:13 Deployment Workshop Overview
34:00 Cloud Shell Setup
35:52 GPU Operator Deployment
43:15 RAG Blueprint Deployment
45:52 Model Deployment Options
46:32 AIQ Model on AKS
47:26 Collection Loading
47:52 Testing & Research Demo
54:02 GPU & System Metrics
56:01 Autoscaling with HPA/KEDA
58:24 Closing Remarks & Speaker Contact
1:00:06 Survey & Wrap‑Up

[eventID:26622]