Rapidata emerges to shorten AI model development cycles from months to days with near real-time RLHF

Source: VentureBeat

Introduction

Despite growing chatter about a future when much human work is automated by AI, one of the ironies of this current tech boom is how stubbornly reliant on human beings it remains—specifically the process of training AI models using reinforcement learning from human feedback (RLHF).

At its simplest, RLHF is a tutoring system: after an AI is trained on curated data, it still makes mistakes or sounds robotic. Human contractors are then hired en masse by AI labs to rate and rank a new model’s outputs while it trains, and the model learns from their ratings, adjusting its behavior to offer higher‑rated outputs. This process is all the more important as AI expands to produce multimedia outputs (video, audio, imagery) that have more nuanced and subjective measures of quality.

Historically, this tutoring process has been a massive logistical headache and PR nightmare for AI companies, relying on fragmented networks of foreign contractors and static labeling pools in specific, low‑income geographic hubs—often cast by the media as low‑wage — even exploitative. It’s also inefficient: AI labs must wait weeks or months for a single batch of feedback, delaying model progress.

Rapidata’s “Gamified” RLHF Platform

A new startup, Rapidata, is making the process far more efficient. Its platform effectively gamifies RLHF by pushing review tasks to nearly 20 million users of popular apps (e.g., Duolingo, Candy Crush). Users can opt‑in to short review tasks instead of watching mobile ads, with data sent back to the commissioning AI lab instantly.

“The platform allows AI labs to iterate on models in near‑real‑time, significantly shortening development timelines compared to traditional methods.” – VentureBeat press release

CEO and founder Jason Corkill added:

“Rapidata makes human judgment available at a global scale and near real‑time, unlocking a future where AI teams can run constant feedback loops and build systems that evolve every day instead of every release cycle.”

Rapidata treats RLHF as high‑speed infrastructure rather than a manual‑labor problem. The company announced an $8.5 million seed round (co‑led by Canaan Partners and IA Ventures, with participation from Acequia Capital and BlueYard) to scale its on‑demand human‑data approach.

The Pub Conversation That Built a Human Cloud

The genesis of Rapidata wasn’t a boardroom—it was a table over a few beers. While a student at ETH Zurich working in robotics and computer vision, Corkill hit the wall that every AI engineer eventually faces: the data‑annotation bottleneck.

“I’ve been working in robotics, AI, and computer vision for quite a few years now, studied at ETH Zurich, and was always frustrated with data annotation,” Corkill recalled. “Whenever you needed humans for large‑scale annotation, the project stopped in its tracks—you could push longer nights, but you had to wait weeks for the annotation work.”

Frustrated by this delay, Corkill and his co‑founders realized that the existing labor model for AI was fundamentally broken for a world moving at the speed of modern compute. While compute scales exponentially, the traditional human workforce—bound by manual onboarding, regional hiring, and slow payment cycles—does not. Rapidata was born from the idea that human judgment could be delivered as a globally distributed, near‑instantaneous service.

Technology: Turning Digital Footprints Into Training Data

The core innovation lies in distribution, not in hiring full‑time annotators in specific regions. Rapidata leverages the existing attention economy of the mobile‑app world:

1. Partnerships with third‑party apps (e.g., Candy Crush, Duolingo).
2. Users are offered a choice: watch a traditional ad or spend a few seconds providing feedback for an AI model.
3. “Hey, would you rather annotate some data instead of watching ads and having companies buy your eyeballs?” Corkill explained.

According to Corkill, 50‑60 % of users opt for the feedback task over a traditional video advertisement. This “crowd intelligence” approach lets AI teams tap into a diverse, global demographic at an unprecedented scale.

Key Metrics

• Global reach: 15 – 20 million people.
• Massive parallelism: 1.5 million human annotations processed in a single hour.
• Speed: Feedback cycles that previously took weeks or months are reduced to hours—or even minutes.
• Quality control: Trust and expertise profiles are built for respondents over time, ensuring complex questions are matched with the most relevant judges.
• Anonymity: Users are tracked via anonymized IDs to ensure consistency and reliability; personal identities are never collected, preserving privacy while optimizing data quality.

Online RLHF: Moving Into the GPU

The most significant technological leap Rapidata enables is what Corkill calls “online RLHF.” Traditionally, AI is trained in disconnected batches:

1. Train the model.
2. Stop.
3. Send data to humans.
4. Wait weeks for labels.
5. Resume training.

This creates a “circle” of information that often lacks fresh human input. Rapidata moves judgment directly into the training loop. Because its network is so fast, it can integrate via API directly with the GPUs running the model.

“We’ve always had this idea of reinforcement learning from human feedback… but you always had to do it in batches,” Corkill said. “Now, because we’re so fast, we have a few clients where the feedback is fed to the model in near‑real‑time, effectively turning RLHF into an online service.”

Bottom Line

Rapidata’s platform reimagines RLHF as a high‑speed, globally distributed service, turning the once‑cumbersome human‑feedback loop into a near‑real‑time infrastructure layer. By tapping into the attention economy of billions of mobile‑app users, it promises to accelerate AI development, improve data quality, and democratize access to human judgment at scale.

Rapidata: Real‑Time Human Feedback for AI Training

Quote from the founder:

“The GPU can calculate an output and immediately request a human in a distributed fashion: ‘I need a human to look at this.’ We get the answer, apply that loss—something that wasn’t possible before.”

Scale of the Platform

• 5,500 + humans per minute provide live feedback.
• Feedback is applied to models running on thousands of GPUs.
• This prevents reward‑model hacking—where two AI models trick each other—by grounding training in genuine human nuance.

Product: Solving for Taste and Global Context

As AI moves beyond simple object recognition into generative media, data‑labeling requirements shift from objective tagging to subjective, “taste‑based” curation.

• Not just “Is this a cat?” but “Is this voice synthesis convincing?”
• Or “Which of these two summaries feels more professional?”

Lily Clifford, CEO of voice‑AI startup Rime, on Rapidata:
“Previously, gathering meaningful feedback meant cobbling together vendors and surveys, segment by segment, or country by country, which didn’t scale. Using Rapidata, we can reach the right audiences—whether in Sweden, Serbia, or the United States—and see how models perform in real customer workflows in days, not months.”

Corkill (Rapidata co‑founder) adds:
“Most models are factually correct, but you’ve received emails that feel… not authentic, right? You can smell an AI email, an AI image, or a video—it’s immediately clear. These models still don’t feel human, and you need human feedback to fix that.”

The Economic and Operational Shift

Rapidata positions itself as an infrastructure layer that removes the need for companies to run their own custom annotation operations.

• Scalable network lowers the barrier for AI teams that previously struggled with the cost and complexity of traditional feedback loops.
• Jared Newman, Canaan Partners (lead investor):

  “Every serious AI deployment depends on human judgment somewhere in the lifecycle. As models move from expertise‑based tasks to taste‑based curation, the demand for scalable human feedback will grow dramatically.”

A Future of “Human Use”

Corkill envisions AI models becoming the primary customers of human judgment—a concept he calls “human use.”

• Example: A car‑designer AI could programmatically call Rapidata to ask 25,000 people in the French market what they think of a specific aesthetic, iterate on that feedback, and refine its design within hours.

“Society is in constant flux,” Corkill notes. “If you simulate a society now, the simulation will be stable for a few months, but then it completely changes because society has evolved differently.”

Funding and Outlook

• $8.5 M in new funding will be used to expand the platform.
• Goal: Make human feedback a real‑time feature, not a bottleneck, as AI scales.

Rapidata aims to be the vital interconnect between silicon and society, providing a distributed, programmatic way to tap into global human brain capacity.