Build Reliable AI Agents Using W&B Training

Learn to build reliable AI agents using Weights & Biases Training's serverless reinforcement learning platform powered by CoreWeave. Discover how to overcome common challenges in productionizing AI applications by leveraging serverless RL, the ART fine-tuning framework, and RULER to fine-tune large language models while improving agent performance and tracking training progress on scalable managed infrastructure. Explore the difficulties of post-training LLMs with reinforcement learning and understand how W&B Training's serverless approach addresses these challenges to make reinforcement learning both reliable and affordable. Follow along with a practical multi-agent contact center example that demonstrates real-world application of these techniques. Master the evaluation of agent LLMs using W&B Weave and gain hands-on experience through a comprehensive notebook walkthrough that covers reinforcement learning code implementation for contact center agents. Learn to view and analyze post-training results in W&B Models and evaluate your post-trained LLM performance in Weave, providing you with the complete toolkit needed to successfully deploy AI agents in production environments.

Syllabus

AI applications are hard to productionize
Post-training LLMs with reinforcement learning
Challenges with reinforcement learning
How W&B Training Serverless RL addresses reinforcement learning challenges
Multi-agent contact center example
Evaluating agent LLMs using W&B Weave
W&B Training quickstart resources
Serverless RL notebook walkthrough
Reinforcement learning code for our contact center agent
Viewing post-training results in W&B Models
Evaluating our post-trained LLM in Weave
Recap, conclusion, and invitation to try the Weights & Biases AI developer platform