In this month’s founder spotlight, our partner Parul Singh sits down with Dereck Paul, cofounder and CEO of Glass Health, a company that is empowering doctors with an AI-powered platform that supports clinical decision-making.

Here is a slightly edited version of the full conversation.

Q: Tell us a little bit about your background and what led you to become a founder.

A: I was interested in becoming a founder even before medical school. While studying biotechnology at Johns Hopkins, I was inspired by professors who were inventors. I tried to start a company as a pre-med student, but it didn’t work out. During medical school at UCSF, I served as the chief of staff at a Y Combinator-backed AI drug discovery company, where I learned the importance of execution, market, and solving real problems. During the COVID-19 pandemic, I saw the weaknesses in our healthcare system and the lack of support for doctors, which led me to co-found Glass with Graham Ramsey. We aimed to provide better software for doctors, and our focus on AI has evolved over the past two years to create products that were previously impossible.

Q: What is your hope about how Glass will help improve healthcare?

A: Many focus on efficiencies and optimization, but what’s special about Glass is our focus on patient outcomes. We care about whether our product improves a patient’s quality of life, wellness, or lifespan and saves lives. As AI models rapidly improve and outperform human experts in many tasks, including medical and clinical reasoning tasks, it seems inevitable that AI models will become the standard of care. This creates a moral imperative for us to build safe, reliable, responsible AI implementations that doctors can use. My biggest hope is that in 5-10 years, we’ll look back and see that we made significant improvements in patient outcomes and saved lives through the application of AI as an assistant to doctors.

Q: How is Glass different from other large language models with healthcare use cases?

A: The biggest difference is that Glass’s large language model application is truly curated by physicians. Physicians are involved in every aspect, from product design and prompt engineering to model selection, validation, fine-tuning, and maintaining the necessary database of information. Unlike applications like ChatGPT or Anthropic’s Claude, which rely on a heterogeneous pre-training process, we need to know exactly where medical recommendations come from. We have a large team of doctors ensuring this accuracy. Additionally, AI clinical decision support is a new product category in healthcare. While AI scribes use speech-to-text models to create useful text from conversations, our focus is on using the emergent properties of large language models and new infrastructure to help providers make complex decisions about diagnosing patients and planning treatments.

Q: What challenges did you face in starting Glass?

A: We’ve been fortunate to be a Y Combinator and Initialized company, receiving a lot of belief and support. However, the challenges of starting Glass included making something out of nothing—building a founding team willing to work without pay, developing the first product version, and acquiring the first 100 users. These initial steps were tough. Then came raising money, scaling the user base, and scaling revenue. Each stage of the startup presented new challenges. Fortunately, we’ve had great investors and mentors helping us overcome these barriers.

Q: You just released a new product last week. Can you tell me more about it? 

A: Absolutely. Glass has been popular among doctors as a standalone web application, introducing them to our technology and integrating it into their workflows. However, embedding healthcare applications into clinician workflows, typically through the electronic health record (EHR), is crucial. This integration allows access to patient data and documentation right alongside the clinician’s work. We just announced and demoed the EHR-integrated version of our platform, set to deploy at one of the largest health systems on the East Coast this summer. This integration can transform the experience of using our product by embedding it in the EHR, right next to patient care, enhancing its usability.

Q: How is your product different from AI scribing products, which also come into the patient conversation with a physician?

A: While both are AI-based, they serve different needs. AI scribe products focus on medical transcription, capturing and summarizing conversations between providers and patients. Our large language model-based application helps providers make complex decisions by analyzing patient data from the EHR, summarizing patient information, and guiding through patient encounters. It goes beyond transcription to assist in interpreting lab results, imaging, and other data, helping create the right plan for the patient. Glass provides AI clinical decision support, a different product category than AI scribes, but they can work well together to serve clinicians.

Q: Any advice you would give to other founders looking to build in healthcare?

A: Building in healthcare is very difficult. Founders at the idea stage see huge opportunities, but if you want quick success, healthcare might not be the best choice. The industry has slow sales cycles, extensive regulatory and legal requirements, and high performance standards for software. Medicine is conservative, so new technology adoption takes time. Building in healthcare requires deep dedication to the problem and users. For me, being close to the problem as a physician and knowing many users personally provides strong motivation. Focusing on how our product can help people keeps me going during tough times. Deep founder-market fit and dedication to solving the problem are crucial for overcoming the many challenges you’ll face.