GeekAcademy Workshop #3 - NVIDIA DLI: Build a Deep Research Agent
About this Course
The "Build a Deep Research Agent" workshop is a self-paced hands-on course designed to teach you how to build and deploy your own deep research agent using NVIDIA's AI-Q Research Assistant Blueprint. Modern AI applications require more than simple question-answering—they need agentic capabilities that can plan, reason, search across multiple sources, and synthesize comprehensive reports. This workshop guides you through deploying a complete enterprise-grade deep research system from the fundamentals up.
Through this workshop, you'll work with NVIDIA's production-ready blueprints to build an autonomous AI system that conducts multi-step research by combining reasoning models, retrieval-augmented generation (RAG), and web search. You'll learn how to deploy NVIDIA Nemotron reasoning models, build a RAG pipeline with vector databases, and orchestrate agentic workflows with human-in-the-loop capabilities. The course emphasizes hands-on learning with immediate application of concepts through Jupyter notebooks and interactive web interfaces.
🎯 Learning Objectives:
By participating in this workshop, you will be equipped to:
Deploy and interact with NVIDIA Nemotron reasoning models using NVIDIA-hosted endpoints.
Build complete RAG pipelines for multimodal document ingestion and semantic retrieval.
Understand agentic AI workflows, including planning, tool calling, and iterative refinement.
Deploy the NVIDIA AI-Q Research Assistant Blueprint for autonomous deep research.
Orchestrate parallel searches across private documents and public web sources.
Generate comprehensive, cited research reports with human-in-the-loop guidance.
✅ Technical Prerequisites:
Basic Python programming experience.
Familiarity with Jupyter Notebooks.
A basic understanding of RAG architecture.
A basic understanding of AI Agent concepts and workflows.
Familiarity with Docker and containerized applications is beneficial.
A general understanding of Large Language Models (LLMs) and embeddings is helpful.