How Ayushi Sinha is Advancing Inclusivity and Scalability in Healthcare AI

AI applications for healthcare have the potential to unlock greater care for more people at lower costs. With US healthcare costs per capita having risen by 46% in the last 10 years, democratizing access to quality care would be transformative. 

But while certain industries have been quick to adopt AI, utilization in healthcare is still in early stages. This is partly because to use AI effectively, the healthcare organizations and technology companies must first develop high-quality patient data sets upon which to train AI applications. 

After growing up in a family of doctors, earning a CS degree from Princeton, and beginning her career in computer vision, Ayushi Sinha recognized that for healthcare AI applications to have a positive impact, the industry would first need to establish diverse and inclusive data sets for model development. So she set out to build Tumerik and help healthcare improve scalability and inclusivity as it transitions into an AI-led era.

Tumerik: Accelerating AI Model Development in Healthcare

Tumerik is building infrastructure that enables AI companies to unlock access to global, diverse patient populations upon which they can train innovative applications for healthcare. 

The healthcare industry is under more pressure than ever before, and organizations have an appetite to explore AI solutions that can help them deliver quality care with greater efficiency. But while AI applications for other sectors can be trained on vast, publicly-available datasets, AI applications for healthcare aren’t able to access training data as easily. 

Thus far, healthcare AI companies have partnered with academic medical centers to acquire access to high-quality, longitudinal, multimodal data sets, often in exchange for company equity. But these arrangements are sub-optimal, as high-growth companies tend to be hesitant to share equity, and academic medical centers tend to have a preference for liquidity over long-term investment bets. Furthermore, medical centers typically have data related to specific patient populations, which can lead to bias in AI models. 

Turmerik is offering an alternative, working with a diverse set of medical organizations across the world to create a comprehensive dataset and making it available to healthcare AI companies. In addition to providing this data, Turmerik offers access to expert physician and specialist data labelers to ensure clinical accuracy, and generates realistic synthetic data to fill gaps and enhance training sets. Turmerik also partners with healthcare organizations and AI companies to help them assess and improve the quality of their datasets, ultimately helping teams across the industry develop more effective AI applications.

Ayushi’s journey to Tumerik

Growing up in a family of doctors that had immigrated from India to rural Tennessee, Ayushi saw first-hand the way in which marginalized communities are often excluded from the large healthcare data sets published by academic centers. She continued to explore this idea while studying Computer Science at Princeton, writing her thesis on data set diversity. 

After college, she began her career as Product Manager at computer vision startup Nines. Building ML products for radiology, she continued to encounter the ways in which healthcare technology was being developed based on data from limited populations. With the rise of LLM applications for healthcare, she knew that developing high-quality, diverse and representative data sets would be key to preventing bias in model outputs. 

Upon receiving her MBA from Harvard, Ayushi set out to launch Tumerik and build a solution that would enable healthcare organizations to utilize AI in a way that ensures scalable and inclusive outcomes. 

Just a year after launching, Tumerik is already partnering with numerous behavioral health and computer vision companies to help them establish a high-quality data foundation for AI use cases.

Ayushi’s advice for fellow founders

Although Ayushi was familiar with many aspects of healthcare before starting Tumerik, she still made sure to investigate the problem space prior to developing the product. In addition to doing discovery calls and user interviews, she took the time to shadow doctors in person and engage in their community. 

“For any founder, make sure you’re a fly on the wall with your users,” Ayushi explains. “A lot of tech founders are told to do discovery calls, but even before you’re asking people questions, watch them and shadow them.”

She notes that this is particularly important when building for audiences beyond developers—professionals who work in specialized environments like hospitals and have complex, industry-specific workflows.

To learn more about Ayushi’s journey, you can follow her on LinkedIn.