The platform leverages a proprietary AI-driven sourcing engine, built by Toptal’s research team, to match specialized experts to training projects instantaneously. By utilizing human experts for RLHF and model evaluation, the company claims it can help organizations scale development with higher accuracy. Prior to this public rollout, the service generated over $5 million in recurring revenue, counting Scale AI and Sterling Labs among its early clients.

Accompanying the launch is the ExpertCrowd-SWE benchmark, designed to stress-test state-of-the-art coding agents. The benchmark spans 87 codebases across 11 programming languages and was constructed entirely by Toptal’s network engineers. According to CEO Taso Du Val, initial tests on models like Opus 5.5 and Fable 5.1 have revealed significant performance deficiencies that other industry benchmarks failed to capture, signaling a new, more rigorous standard for evaluating AI coding capabilities.