iMerit Champions Quality Data Over Quantity for the Future of AI

The AI data platform iMerit claims that the essential factor for enhancing AI integration at the enterprise level is the improvement of data quality, not just the increase in data quantity. The firm emphasizes that quality data originates from domain experts, rather than numerous gig workers, in fields such as mathematics, medicine, healthcare, finance, and autonomy.

“Recruiting and retaining top cognitive specialists is vital as we need to tailor these large models to meet specific enterprise AI challenges,” Radha Basu, CEO and founder of iMerit, mentioned in an interview with TechCrunch.

In the past nine years, this startup, based in California and India, has positioned itself as a trusted data annotation partner for companies in sectors like computer vision, medical imaging, and autonomous mobility, demanding accurate human-in-the-loop labeling.

iMerit is launching its Scholars program, moving out of beta, exclusively shared with TechCrunch. This initiative aims to build a team of experts dedicated to enhancing generative AI models for enterprise applications and foundational models.

The company’s clientele includes leading AI firms, among them three of the seven major generative AI businesses, eight key autonomous vehicle companies, three significant U.S. government agencies, and two of the top three cloud service providers.

This development comes in the wake of Scale AI’s recent turmoil, losing founder and CEO Alexandr Wang to Meta, which has acquired a 49% stake in the company. As a result, many of Scale’s major clients, including Google, OpenAI, Microsoft, and xAI, have reduced their engagements due to fears that Meta might gain access to their product strategies.

iMerit does not position itself as a substitute for Scale AI’s primary offering of high-volume, developer-centric “blitz data.” Instead, it is seizing this chance to highlight the importance of expert-driven, high-quality data, marked by deep human insight and domain-specific guidance.

“We see ourselves as the responsible adults in the room,” stated Rob Laing, iMerit’s VP of global specialist workforce, to TechCrunch. “Substantial financial investments are currently being made into AI. Various innovators are creating extensive human workforces. However, the output from this mass approach often fails to meet the quality standards enterprises demand.”

Basu mentioned the increasing numbers of healthcare scribes emerging from foundational large language models.

“Without the expertise of a cardiologist or physician, what you’re likely to produce achieves around 50% or 60% accuracy,” Basu remarked. “Our goal is 99% accuracy. We seek to challenge and refine the model. Expert-led AI can accomplish this level of precision for enterprises.”

iMerit’s specialists focus on refining or “tormenting” enterprise and foundational AI models using the company’s proprietary platform, Ango Hub. This platform allows iMerit’s “Scholars” to engage with the customer’s model to create and evaluate challenges the model can solve.

For iMerit, the recruitment and retention of cognitive specialists are crucial, as these experts participate in multi-year projects instead of merely performing short tasks. The company boasts a 91% retention rate, with women constituting 50% of its expert workforce.

Laing, whose background in founding the human translation platform myGengo has given him insights into crowdsourcing dynamics, emphasized that attracting individuals for routine tasks is easy. However, fostering a sense of community requires a more personalized approach.

“When someone joins the Scholars program, they do not become just a name in a database; they actually meet the team,” Laing elaborated. “They participate in collaborative discussions and are inspired to deliver their best work. We maintain high selectivity in our recruitment process.”

“In the years to come, I believe organizations like iMerit, which focus on engagement, retention, and quality, will become the preferred partners for AI training,” Laing added.

Currently, iMerit collaborates with over 4,000 Scholars and aims to grow this number as it expands. Basu informed TechCrunch that even though the company has not raised funds since 2020, when it attracted investors like Khosla Ventures, Omidyar Network, Dell.org, and British International Investment, it remains sustainable and profitable. With its own financial resources, the company is capable of scaling up to 10,000 experts, though future growth may require external funding, which iMerit is open to considering but not urgently pursuing.

In the past year, iMerit has primarily focused on Scholars in the healthcare sector but plans to expand into other enterprise areas like finance and medicine. Laing pointed out that generative AI is its fastest-growing segment, as leading AI companies partner with iMerit to enhance their foundational models.

“The amount of readily available data on the internet has decreased, and lower-level human input data has also become commoditized,” Laing commented. “The current focus is on fine-tuning these models to strive for AGI or superintelligence.”