Olam Labs YC 申请

Olam Labs

Multi-agent sims for model evals and training.

We're building multi-agent simulation environments for evaluating and training AI agents in ways that go beyond what standard datasets allow. Our main product is Multi-Agent Arena, where humans and multiple AI agents play social strategy games. Researchers use it to test and train models on complex behaviors that are hard to measure elsewhere. We also generate real-world data on agentic performance and socialization.

We're based in San Francisco, CA, and will remain here after YC.

San Francisco is where much of the AI research community is located. We're close to collaborators and early users, and being in SF helps us stay plugged into the ecosystem.