Observing and Improving AI Agents with Langfuse: a hands-on workshop

BUILD LLM APPS THAT ACTUALLY WORK IN PRODUCTION

Whether you're shipping a production AI product or building something ambitious over five days, the teams that move fastest are the ones who can measure what's working. This workshop gives you the observability and eval workflow to iterate with confidence, not instinct.

THE PROBLEM?

Most LLM apps are built blind. Prompt changes, model swaps, tool call tweaks, teams iterate fast but measure nothing. You can't tell if your last change made the agent better or worse until something breaks in front of a user.

That's the bottleneck between a prototype and something that holds up, whether that's a Demo Day pitch or a production deployment.

WHY JOIN?

Langfuse is your unfair advantage. While other teams guess why their agent broke, you'll have full trace visibility, versioned prompts, and an eval loop that tells you in minutes whether your last change helped or hurt. That's the difference between a demo that impresses and one that just about holds together.

From first trace to full eval workflow in 2 hours:

Full trace visibility across every agent step and tool call

Prompts versioned and deployed outside your codebase

Monitoring that catches failure modes automatically

A dataset + experiment run you can reuse on every future iteration

WHO THIS IS FOR?

AI developers building LLM applications who want a structured, repeatable way to improve their apps

Builders who want to move faster than every other team

Tech leads evaluating observability tooling before taking an AI product to production

You should be comfortable reading TypeScript. No prior Langfuse experience needed.