Agent Evals: The Truth Machine w/ Composio, Braintrust, Fireworks, and Replit

​How leading AI teams test, trace, and trust agents in production?

​This is the go-to panel for builders who care about evaluating agents as they are heading into production, not just impressive demos.

​Join Fireworks, Braintrust, Replit, and Composio for a conversation on agent evals, reliability, observability, and the new infrastructure needed to make AI agents trustworthy in the real world.

​Time & Location

​Time: June 23rd Tuesday, 6:30PM - 8:30 PM

​Location: Notion HQ, SF

​Featuring

• ​Ankur Goyal, CEO of Braintrust

• ​Soham Ganatra, CEO of Composio

• ​Dima Dzhulgakov, CTO of Fireworks AI

• ​Zhen Li, Engineer lead at Replit, Creator of Replit Agent

​Cohosts

​Braintrust
Braintrust is the AI observability and evals platform for teams building production AI systems. It helps builders trace real-world behavior, run evaluations, and catch regressions before they reach users.

​Composio
Composio is the agent action harness, empowering agents to take actions on 1000+ apps, 10000+ tools, APIs, auth, and real-world workflows. It helps developers move agents from isolated demos to systems that can actually take reliable action.

​Fireworks AI
Fireworks AI provides fast, scalable infrastructure for serving, fine-tuning, and deploying generative AI models. For agent builders, Fireworks helps teams optimize model performance, latency, and reliability in production environments.

​Replit
Replit is an AI-powered software creation platform where developers and non-developers can build, ship, and iterate on apps quickly. Its work on coding agents makes it a key voice in understanding how agents plan, write code, use tools, and recover from mistakes in real-world development.