TL;DR
- iMerit, a U.S.–India-based AI data platform, argues that quality, not quantity, will define the next era of AI training data.
- The company has launched its Scholars program to scale a vetted workforce of cognitive experts across verticals like healthcare, finance, and autonomy.
- As competitors like Scale AI see client attrition after Meta’s acquisition, iMerit is gaining trust from three of the “Big 7” GenAI firms and multiple U.S. agencies.
- iMerit claims 91% expert retention, 50% of whom are women, and runs a proprietary platform called Ango Hub for model refinement.
- CEO Radha Basu says iMerit is profitable, sustainable, and poised to scale up to 10,000 experts using internal reserves.
Rethinking the Data Pipeline: From Gig Workers to Domain Experts
As enterprise adoption of AI accelerates, iMerit believes the era of simply accumulating more training data is coming to an end. Instead, the future lies in data precision, tailored by experts with deep domain knowledge rather than crowdsourced labor.
“You want to question the model. You want to break it. You want to fix it,” said Radha Basu, CEO and founder of iMerit.
Founded nearly a decade ago, iMerit has steadily built a reputation in human-in-the-loop data annotation for mission-critical fields like medical imaging, autonomous mobility, and financial intelligence. Now, it’s bringing its flagship Scholars program into the spotlight—designed to create a workforce of cognitive experts who fine-tune AI systems with deep contextual understanding.
iMerit vs the Competition
| Metric | iMerit |
| Headquarters | California, USA and Kolkata, India |
| Founded | 2012 |
| Scholars in Workforce | 4,000+ |
| Retention Rate | 91% |
| % of Women Experts | 50% |
| Investors | Khosla Ventures, Omidyar Network, Dell.org, British International Investment |
| Core Clients | 3 of top 7 GenAI companies, 3 U.S. government agencies, 8 AV firms, 2 of top 3 cloud vendors |
| Proprietary Platform | Ango Hub |
| Last Raised Funding | 2020 |
| Revenue Model | Sustainable, profitable; scaling via internal reserves |
The Scholars Program: High-Touch, High-Retention Workforce
The iMerit Scholars program is positioned as a response to AI’s increasing complexity. With AI tools penetrating sectors that cannot afford low accuracy—like cardiology, finance, and public sector compliance—the margin for data error is shrinking.
Unlike gig-style annotation services, the Scholars program emphasizes long-term project engagement, with human feedback loops that fine-tune foundation models and enterprise-specific LLMs.
“Instead of someone being a name on a database… they actually meet folks on the team,” said Rob Laing, VP of Global Specialist Workforce and former co-founder of translation platform myGengo.
Responding to Industry Shifts: Scale AI’s Decline, iMerit’s Rise
The timing of iMerit’s rise is notable. As Scale AI, once the most prominent name in data labeling, faces fallout from Meta’s 49% acquisition, major clients such as Google, OpenAI, and Microsoft are reportedly pulling back over privacy concerns.
iMerit, by contrast, is seen as a neutral and secure alternative, not aligned with any competitive LLM entity.
“We’re the adults in the room,” Laing remarked. “The mass-market approach isn’t producing the level of quality that enterprise-grade AI now demands.”
Ango Hub: The Engine Behind the Expertise
iMerit’s proprietary Ango Hub platform powers the Scholars program. It enables experts to challenge models, design evaluation sets, and probe weaknesses—a process Basu refers to as tormenting the model.
This active evaluation loop helps refine AI systems at a much higher fidelity than conventional annotation tools. It’s not just about labels—it’s about questioning assumptions, surfacing edge cases, and optimizing confidence thresholds.
Serving Generative AI, Government, and Healthcare
Among iMerit’s clients are three of the top seven generative AI developers, as well as U.S. federal agencies, and eight major AV startups. The healthcare vertical has proven particularly important as demand for medical scribes, diagnostic tools, and clinical assistant LLMs has surged.
“If you don’t have the expertise of the physician, you’re just creating something that’s maybe 60% accurate,” Basu explained. “You want that to be 99%.”
With generative AI quickly becoming a critical layer across enterprise operations, iMerit’s human-led fine-tuning is becoming an indispensable layer of assurance.
Sustainable Scale Without Outside Pressure
iMerit has not raised fresh capital since 2020, and yet the company says it is profitable and cash-flow positive. Basu affirms that iMerit has the internal capacity to scale from 4,000 to 10,000 experts without seeking new capital.
“To scale beyond that, we’re open to investment—but we’re not desperate for it,” she said.
This self-sufficiency positions iMerit to scale deliberately, prioritizing expert retention and data quality over market land grabs.
AI’s Next Stage: Quality, Not Quantity
According to Laing, much of the “free data” from the open internet has already been scraped. And lower-tier human annotation work is increasingly commoditized by automation and synthetic generation.
“What’s next,” Laing argues, “is tuning for AGI and superintelligence. That’s where curated, expert-generated data becomes the competitive edge.”
In that light, iMerit’s expert-led, mission-critical labeling model may prove to be the new foundation of trustworthy AI infrastructure.
