Entrepreneur, CEO, And Co-Founder of Hippocratic AI

CEO Munjal Shah on How Hippocratic AI Is Training ‘Empathetic’ Healthcare LLMs

Hippocratic AI focuses on leveraging large language models (LLMs) to engage in long, multiturn voice conversations within healthcare, distinguishing its approach from other AI applications that prioritize tasks like question answering. The company aims to enhance healthcare delivery by developing empathetic, nondiagnostic AI agents that support healthcare providers through patient conversations. Central to this effort is Polaris, a unique “constellation architecture” featuring a primary AI agent complemented by specialist support agents. These agents manage specific tasks, such as medication management and lab result analysis, ensuring safety and accuracy through a layered system of checks and balances.

Training these AI agents involves a multi-stage process, beginning with the integration of extensive medical data and simulated dialogues with patient actors, designed to instill empathy and effective communication skills. Real-world feedback from thousands of healthcare professionals plays a critical role in refining the models. The company is cautious about scaling its technology, acknowledging the challenges of maintaining empathetic and coherent communication across millions of patients. By working alongside human healthcare professionals, Hippocratic AI aims to augment, not replace, human-led care, with the goal of improving patient outcomes across the healthcare system.

Read full article

more news

MUNJAL SHAH’S AUDACIOUS BET ON THE FUTURE OF HEALTHCARE AI

As co-founder and CEO of Hippocratic AI, Munjal Shah is developing artificial intelligence that he believes could fundamentally reshape how medical care is delivered. His company’s generative AI large language models, purpose-built for healthcare, aim to take on countless routine tasks. Those tasks are currently performed by healthcare providers, everything from preoperative instructions to chronic disease management check-ins.

“What if instead of doing a co-pilot model, we do autopilot?” Shah posits, contrasting his vision with AI tools designed merely to assist human clinicians. “What if we build fully automated AIs that call people on the phone and talk to them? Imagine an AI that can do nondiagnostic, low-risk tasks like preoperative calls and medication reminders?”

Read more >