Zibra Labs builds distributed compute clusters that leverage the most economical CPUs and GPUs across hyperscalers and neoclouds for AI workloads. The company initially targets quantitative trading firms with large-scale high-performance computing clusters for backtesting and parallel simulation, with applications extending to post-training reinforcement learning, fine-tuning, long-horizon agents, and batch inference. It was founded by Ibrahim Rabbani, a former tech lead of Ray who previously built databases at LinkedIn, and Zac Policzer, who has extensive experience with large-scale databases. Zibra Labs is based in San Francisco and is part of Y Combinator's Spring 2026 batch.