Pioneering Diversity in Healthcare AI: A Conversation with Ayushi Sinha of Turmerik AI
As part of our new interview series highlighting our incredible members, Women Who AI spoke with Ayushi Sinha, founder of Turmerik AI, a startup that's revolutionizing clinical trials through AI while championing diversity in healthcare data.
---
The Mission: Health Equity Through Data Diversity
"Clinical trials today really lack representation of women and people of color, and this has major ramifications for health equity," explains Sinha.
Consider this startling fact: The widely accepted recommendation that people need eight hours of sleep was based on studies of white men in college. When researchers finally conducted the same study with women, they discovered women actually need nine hours on average.
"Imagine how our entire workday is designed around the assumption that people need eight hours of sleep, which isn't true for half the population," she points out.
This represents just one example of how the lack of diversity in medical research can have far-reaching consequences for healthcare outcomes and everyday life. It's precisely this problem that Turmerik AI is addressing.
The Solution: AI-Enabled Global Networks
Turmerik AI connects international hospitals with diverse patient populations to biopharma companies looking to run clinical trials. Their approach is powered by a comprehensive AI suite that:
- Digitizes paper records
- Generates IRB paperwork and documentation
- Uses AI algorithms to match patients to appropriate trials
- Leverages advanced LLMs to parse unstructured data at scale
"We're very end-to-end because we realized we have to augment doctors. We can't add anything to their workload," Sinha explains. This practical approach has allowed Turmerik AI to build a network spanning hospitals in India, Thailand, Israel, and Canada.
Future Directions: Specialized vs. General AI Models
As Turmerik AI continues to grow, Sinha is exploring important questions about the future of AI in healthcare.
"The verdict isn't clear," she says when asked whether general AI models work as well as healthcare-specific ones. "All of these models are trained on public and open-source data, which is great. But we're thinking a lot about how to use federated learning to help hospitals share their data securely."
With 70% of data going into medical AI products coming from just three states, Turmerik AI is positioning itself to help reshape how healthcare AI models are built and trained.
This research could help determine whether healthcare AI needs specialized models trained on diverse, representative data or whether general models can be effectively adapted to the healthcare domain.
Navigating Regulatory Hurdles with Trust
The healthcare industry presents significant regulatory challenges for AI startups. As Sinha puts it, "You can't be a 'random AI girly' in healthcare—you need anchors of credibility." Her strategy for building trust and navigating these hurdles has been multifaceted. She's leveraged Key Opinion Leaders (KOLs), partnering with respected MDs who co-design tools and vouch for their efficacy.
Turmerik AI has built hybrid teams that mix technical experts with medical advisors to balance innovation and compliance. They've also taken a phased approach to compliance, starting with clinical trial documentation automation before expanding to more complex patient matching algorithms. "I'm not a doctor, but people on our team are doctors, and our customers are well-known doctors," she explains. "I'm standing on the shoulders of giants."
Her Founder's Journey: Bridging Computer Science and Healthcare
Coming from a family of doctors and growing up in rural Tennessee, Sinha developed an early awareness of health equity issues. While her family members pursued medical careers, she took a different path:
"I wanted to focus on scaling access to healthcare, so I went the CS route," she explains. At Princeton, she even wrote her thesis on dataset diversity—technical work that would later become fundamental to Turmeric's mission.
After gaining experience building FDA-approved AI algorithms for radiology at a previous startup, she launched Turmerik AI during her time at Harvard Business School. The business idea evolved from an expert data labeling company she ran during her MBA, which gave her "free access to amazing CEOs and heads of hospital systems" through which she discovered the needs in the clinical trial space.
Connect with Turmerik AI
Interested in exploring dataset diversity for medical AI? Turmerik AI is actively seeking connections with AI companies facing data challenges in the healthcare space. They're eager to understand how they might help build more accurate models through robust, longitudinal, multimodal datasets. Connect at [email protected].