iMerit: The Future of AI Is Better Data, Not Just More Data
Expert-Led Annotation Emerges as the Key to High-Quality, Enterprise-Ready AI
Shifting the Focus: Quality Over Quantity in AI Data
AI data platform iMerit is betting big on a new direction for artificial intelligence—prioritizing better-quality data over simply amassing more data.
- The company argues that the next wave of enterprise AI will depend on deep domain expertise, not mass labeling by crowdsourced gig workers.
- According to CEO and founder Radha Basu, “What’s become exceedingly important is the ability to attract and retain the best cognitive experts… to make these large models very customized towards solving enterprise AI problems.”
Building an Expert Workforce with iMerit Scholars
Over nine years, iMerit has grown into a trusted data annotation partner for industries requiring high-accuracy, human-in-the-loop labeling, including computer vision, medical imaging, and autonomous vehicles.
- Now, iMerit is officially launching its Scholars program, aimed at building a workforce of domain experts—mathematicians, physicians, finance professionals, and more—to finetune generative and foundational AI models for enterprise use.
Notable Clients & Impact
- iMerit’s client list includes three of the “big seven” generative AI companies, top autonomous vehicle firms, major U.S. government agencies, and leading cloud providers.
- Scholars aren’t just short-term contributors: 91% retention rate, with half of experts being women, shows a strong, engaged talent pool.
The Industry Landscape: Why Expert Data Matters Now
The timing is significant. Competitor Scale AI recently lost its CEO and saw major clients pull back after Meta’s investment.
- iMerit is betting that deeply vetted, expert-led annotation—not the high-speed, crowdsourced approach—is what’s needed for the next leap in AI.
- “The output that they’re getting from that mass approach and that very quick speed to market… is not at the level of quality that enterprises need,” says Rob Laing, VP of global specialist workforce at iMerit.
Why Accuracy Is Crucial
- In fields like healthcare, using only generic data or non-expert annotation can lead to models that are “maybe 50% or 60% accurate,” Basu notes.
- For enterprise applications, especially in regulated industries, accuracy must approach 99%—requiring expert judgment, continuous model evaluation, and robust challenge-testing.
Human-in-the-Loop: A New Standard for AI Training
iMerit’s proprietary Ango Hub platform enables its Scholars to finetune, “torment,” and evaluate AI models, generating and assessing real-world problems for models to solve.
- The focus is on long-term engagement and community, not anonymous gig work.
- Scholars meet regularly, collaborate closely, and are highly selective recruits—ensuring consistently high standards.
Sustainable, Profitable, and Scaling
iMerit operates with over 4,000 Scholars and plans to grow to 10,000 experts using its own resources.
- The company has been profitable and hasn’t needed new funding since 2020, thanks to strong investor backing and steady growth.
- Expansion beyond healthcare to finance, medicine, and more is underway, as demand for expert-tuned generative AI rapidly rises.
The Road Ahead: Quality Data Is the Differentiator
With “free” data and generic human labeling now commoditized, the race is on to provide high-quality, expert-driven data for AI models targeting AGI and superintelligence.
- “Companies like iMerit that are really focusing on engagement, retention, and quality are going to be the go-to partners for training the next generation of AI,” Laing emphasizes.
