Rapidata launches platform cutting AI cycles from months to hours

Lead

Rapidata, a Zurich-based startup, announced this week a platform that lets AI labs shorten model training cycles, cutting development time from months to hours by integrating near-real-time human feedback. The company raised a US$8.5 million seed round led by Canaan Partners and IA Ventures, with participation from Acequia Capital and BlueYard. The system connects AI evaluation tasks to a global audience of up to 20 million users of major apps like Duolingo and Candy Crush through quick, voluntary missions.

Development Section

In AI, training with human feedback—known as RLHF (reinforcement learning with human feedback)—is key to tuning models like chatbots or image generators, making them less robotic and more human-like. Traditionally, this relies on fragmented contractor networks often based in low-income countries, leading to criticism over exploitation and long delays: feedback rounds can take weeks or months.

Rapidata aims to turn RLHF into a high-speed digital infrastructure. Instead of fixed evaluators, the startup taps into the attention of global users of popular apps, offering them the option to review AI outputs instead of watching ads. Founder Jason Corkill says between 50% and 60% of users choose these feedback tasks.

With this model, Rapidata processes up to 1.5 million annotations per hour, with about 5,500 people per minute providing real-time evaluations. This enables AI projects to receive human feedback directly in the training loop via GPU integration—specialized AI processing chips—eliminating batch delays. The system uses anonymous IDs to protect respondent privacy and builds trust profiles over time, assigning more complex tasks to qualified users.

Beyond speed gains, the demographic diversity of evaluators improves model quality, especially for subjective tasks like judging how natural an AI-generated voice sounds or which text summary seems more professional. Companies like Rime, an AI voice firm, report testing models across countries and real workflows in days, instead of months of negotiations and regional targeting.

Outlook and Perspectives

The impact on AI is significant: Rapidata removes the need for in-house annotation operations, cuts costs, and enables near-continuous update cycles, becoming essential infrastructure for startups and large labs. Jared Newman of Canaan Partners notes scalable human feedback demand will grow as AI moves from objective tasks to curation based on taste and cultural context.

Looking ahead, AI models are expected to request feedback programmatically from Rapidata’s “human cloud,” testing products or concepts with tens of thousands of people within hours, not weeks. Founder Jason Corkill says the platform makes the human element a real-time resource, aligning society’s pace with algorithm evolution.

With the new US$8.5 million funding, Rapidata plans to expand its global base and solidify its role as a bridge between data processing and human judgment, making societal feedback an active, instant part of the AI development cycle.