Serial entrepreneur Munjal Shah has launched a new health tech startup called Hippocratic AI that aims to leverage large language models (LLMs) to provide personalized chronic care and patient support services. With 68 million Americans suffering from multiple chronic conditions but only a few hundred thousand specialized nurses available, Shah sees AI as a way to expand access to caretakers greatly.
Hippocratic AI plans to develop LLMs uniquely trained on medical datasets to assist patients with non-diagnostic needs. This could include medication and appointment reminders, diet tips, specialist recommendations, and administrative support for tasks like medical billing. Shah believes LLMs can compile, synthesize, and communicate medical knowledge to augment overburdened nursing staff. The goal is not to replace nurses but rather “super staff” to improve outcomes.
From E-Commerce AI to Healthcare LLMs
Shah has a strong track record in AI, having founded startups utilizing machine learning for e-commerce acquired by Alibaba and Google. However, those systems relied on what Shah calls “classifier AI,” focused on categorizing products or data points based on rules.
LLMs like ChatGPT have more open-ended generative abilities. Their neural networks can create unique written responses tailored to specific patient needs and communicate empathetically. Seeing the potential to apply this technology to healthcare’s provider shortage crisis, Shah launched Hippocratic AI.
Avoiding Diagnosis But Still Adding Value
Munjal Shah recognizes that while promising, LLMs still make concerning mistakes. Allowing an LLM to directly diagnose conditions or dictate treatment could put patients at risk, so Hippocratic AI avoids those use cases. However, Shah sees plenty of impactful, safe applications for non-clinical support.
For instance, LLMs could field routine calls about billing questions and clarify insurance policies more in-depth than most staff have time for. They can call patients with expected test results rather than occupying a nurse. LLMs may also excel at preventative health guidance, leveraging the latest medical research to offer personalized chronic disease management tips.
Training Responsibly on Healthcare Datas, Hippocratic AI’s LLM will undergo specialized training on peer-reviewed medical literature and documentation of insurance documentation to minimize mistakesion. This deep domain-specific grounding aims to prevent “hallucinated” false information. The LLM’s responses will also be vetted by doctors and nurses working in the areas it’s designed to augment.
Shah summarizes his motivation as wanting to provide every patient with dedicated individual caretakers – even if it requires AI to make that a reality. With nursing shortages growing, responsible applications of generative LLMs may help democratize chronic care support. Hippocratic AI provides an intriguing case study for the life-changing potential of AI in health.
