The Weights & Biases AI agent workshop is your chance to join top AI engineers for a day of agent building and model training. We'll help you reach the next level in AI development, where agents self-improve using Weights & Biases tools.
First learn how to use reinforcement learning (RL) to improve the reliability of LLMs on agentic tasks. Then build an agent harness around this model for continuous learning. You’ll walk away with a clear understanding of agent self-improving workflows and how to apply them to your use case.
Who Should Attend: AI Engineers / ML / AI Researchers
What to Expect From the Program:
Build and evaluate an open-source LLM-powered agent on real-world tasks
Design structured tasks and evaluation pipelines to prevent model drift and enable cost-efficient, self-improving agents
Architect production-ready agents by adding tool use, MCP integration, skills frameworks, and iterative evaluation workflows
Session One
ART-E-Mini: Specialize open source models through reinforcement learning
In this session, you’ll learn how to continuously improve an AI agent as it works with end users using reinforcement learning. You’ll build an agent that searches real emails with an open-source model, see how it performs and then make it better with serverless RL on powerful CoreWeave GPUs. We’ll walk through how to define tasks to prevent model drift and enable incredibly cost-effective agents with Weights & Biases tools.
Session Two
From brains to body: Supercharging your specialized agent
In this session you’ll learn how to get your newly specialized model to work efficiently and equip your model with the skillset needed to become a powerful agent. We’ll explore agent skills, tool use, MCPs, evaluations, and more. Throughout the session, you’ll gain a deep understanding of how your agent can be improved using tools such as W&B Weave.
