Zurich-based Rapidata has raised €7.2 million ($8.5 million) in seed funding to eliminate one of the primary bottlenecks in modern AI development: the slow pace of human labeling. The round was co-led by Canaan Partners and IA Ventures, with participation from BlueYard and Acequia Capital.
The big picture: While model architectures and compute power have surged, Reinforcement Learning from Human Feedback (RLHF) remains a manual, week-long struggle. Rapidata addresses this by distributing micro-tasks through a global network of consumer applications, reaching tens of millions of users daily to compress feedback cycles from months into a single day.
How it works:
- Scalable RLHF: Provides high-volume human judgments for training foundation models in voice, motion, and text.
- Rapid Iteration: Integrates directly into developer workflows to allow constant, real-time model improvements rather than release-cycle updates.
- Global Reach: Leverages crowd intelligence across diverse geographies to test models in real-world customer contexts.
The catch: While Rapidata claims to offer “near real-time” human judgment, the quality of feedback from opt-in tasks within consumer apps is notoriously difficult to maintain. Using “taste-based curation” from a general audience may work for simple preferences, but it risks failing for the high-stakes, expertise-heavy tasks required by clinical documentation or specialized engineering models. Furthermore, as the industry moves toward “AI-judging-AI” (RLAIF), Rapidata must prove that human “taste” remains a unique, non-replicable advantage that justifies the operational cost over fully automated synthetic evaluation.
Key Details
- Funding: €7.2M (Seed)
- Lead: Canaan Partners, IA Ventures
- CEO: Jason Corkill
- Sector: AI Infrastructure / Data Labeling
