Ragas Team Emerges from Stealth as Vibrant Labs with $2.5M pre-seed round to Build Simulation Environments for Long-Horizon Agents

We set out two years ago to improve AI systems. Inserting evaluations into the workflow to catch errors in development and not in production. We were accepted into YCombinator on that premise and received a $2.5M pre-seed round to build whatever comes next. 

We saw that AI systems were evolving from being primarily used for interaction, i.e. Q&A chat, to automating systems that exhibit reasoning,taking actions on behalf of humans and automating parts of the human workforce through this. This requires a new approach, both evaluating its capabilities and training what's possible next through reinforcement learning. 

Today, we are coming out of stealth as Vibrant Labs. We are tackling the next problem of evaluating and enabling training of long-horizon agents using simulated environments. Not just evaluations, but creating complete Reinforcement Learning (RL)-ready worlds where agents can fail safely, learn new skills, and reason and improve.

After we built evaluators (verifiers) for various types of AI systems, we saw that those can now be used to improve AI through new training paradigms like RLVR. Now, agents need environments where they can learn, fail, and improve before they touch real systems. The traditional way of evaluating and training AI agents has hit a wall. Simulations are the future.

About Vibrant Labs
Founded by Shahul ES and Jithin James, the creators of Ragas, Vibrant Labs builds controlled environments for evaluating agentic AI systems. Backed by YCombinator, Heavybit, PioneerFund, and Hike Ventures. 

Learn more at vibrantlabs.com.