NERSC, in collaboration with NVIDIA and the OpenACC organization, will host an End-to-End AI for Science bootcamp on April 23 - 24, 2026.
Agentic AI employs sophisticated reasoning to solve complex problems, utilizing data from multiple sources to analyze, strategize, and complete tasks independently. There are many powerful building blocks that developers and researchers can use to enable and accelerate the development and deployment of their agentic AI workflows.
The Agentic AI Bootcamp covers some of these powerful tools, including Model Context Protocol (MCP), an open protocol providing a standardized method for connecting AI models to various data sources and tools; NVIDIA NIM, a set of easy-to-use microservices that provides pre-optimized models and industry-standard APIs for building powerful AI agents, co-pilots, chatbots, and assistants; and LangGraph, an open-source library for complex and stateful workflows for AI agents.
Attendees are guided through the creation of MCP servers based on Stdio/HTTP using the high/low level MCP SDKs, building MCP clients to interact with MCP Servers, and using MCP Clients/Servers with NVIDIA NIM in LLM/Agentic workflows. Attendees will then apply these learned skills in a hands-on challenge to replace parts of an existing workflow with MCP.
This virtual bootcamp is a hands-on learning experience where you will be guided through step-by-step instructions with teaching assistants on hand to help throughout.
We will be using Perlmutter GPUs for the Bootcamp. This event has limited capacity, so we recommend applying early. Training accounts for Perlmutter will be provided for non-NERSC users.
Prerequisites
Basics of Python and Python package manager (such as pip, conda, or uv)
Apply to Attend
For detailed information on how to apply, please refer to the Open Hackathons’ Bootcamp Event Page. The application deadline is April 1, 2026. This event has limited capacity. Please apply early, and note that acceptance is not confirmed until you have received a confirmation email.
Agenda (Tentative)
Day 0: April 22
| Time | Topic |
|---|---|
| 10 - 11 a.m. | Cluster Dry Run |
Day 1: April 23
| Time | Topic |
|---|---|
| 9 - 9:30 a.m. | Introduction and Cluster |
| 9:30 - 10:30 a.m. | Lab 1: Using NVIDIA NIM via Cloud and Local Endpoints |
| 10:30 - 11:15 a.m. | Lab 2: Introduction to Model Context Protocol (MCP) |
| 11:15 - 11:45 p.m. | Lab 3: Low-Level MCP Server Implementation |
| 11:45 - 12:15 p.m. | Break |
| 12:15 - 1 p.m. | Lab 4: Building Agentic Workflows with LangGraph |
| 1 - 1:30 p.m. | Challenge Description and Wrap-Up |
| 1:30 - 2:30 p.m. | Project Discussions |
Day 2: April 24
| Time | Topic |
|---|---|
| Asynchronous Assignment: Challenge | |
| Challenge: Enhancing a Music Store Agent Workflow with Model Context Protocol (MCP). Participants will work on: | |
| - MCP Server Development utilizing the low-level Python SDK. | |
| - Reusable Server Packaging for easy integration and reuse by MCP clients. | |
| - Design and implement workflows that incorporate human-in-the-loop interactions via LangGraph. | |
| - NVIDIA NIM Integration as LLM backbone in agentic workflows |