Rapidata's $8.5M Seed Round Turns Mobile Game Users Into Real-Time AI Trainers
The Development
Zurich-based Rapidata emerged from stealth on February 19, 2026, announcing an $8.5 million seed round co-led by Canaan Partners and IA Ventures, with participation from Acequia Capital and BlueYard. The company has built a platform that reroutes mobile ad inventory toward AI training tasks: users inside apps like Duolingo and Candy Crush are offered the choice to complete short AI feedback tasks instead of watching video ads. Between 50-60% of users opt in. The result is a network of 15-20 million annotators capable of processing 1.5 million human annotations per hour—reducing RLHF cycles that traditionally take weeks to hours or minutes. Critically, Rapidata has integrated directly into GPU training loops via API, enabling what CEO Jason Corkill calls 'online RLHF': live human feedback applied to model outputs mid-training, in real time, at a rate of roughly 5,500 human responses per minute.
Our Take
Traditional RLHF is a procurement problem masquerading as a technical one—slow vendor networks, regional contractor pools, batch delays. Rapidata dissolves that by hijacking the one asset already distributed globally at scale: mobile attention. What's easy to miss is what this does to the ad impression. Rapidata isn't just offering an alternative to video ads; it's asserting that human cognitive engagement has a higher-value buyer than CPG brands running pre-roll. If 50-60% opt-in rates hold at scale, publishers and ad networks will eventually have to answer why they're still selling eyeballs to the lowest bidder when AI labs will pay for actual judgment. The 'online RLHF' capability—feedback injected directly into GPU training loops—is the real long-term lever. It transforms human input from a periodic input into continuous infrastructure, which shifts the entire value chain of AI development.
The Exploit
Action Item
If your brand runs mobile ad campaigns in casual gaming or language apps, audit your Candy Crush and Duolingo inventory allocations now—Rapidata's opt-in model will compress available ad impressions in those placements as publisher adoption scales through Q2 2026.