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?”
It’s an audacious vision, one that’s attracted $120 million in venture funding and a $500 million valuation for the young startup. But Munjal Shah is no wide-eyed newcomer to either technology or entrepreneurship. He’s on his fourth major venture, having already notched two successful exits, including the sale of visual search engine like.com to Google in 2010 for a reported nine-figure sum.
Building Safe Specialists
What sets Hippocratic AI apart from others in the fast-growing GenAI space, Shah argues, is its laser focus on healthcare applications and an uncompromising commitment to patient safety. While general-purpose LLMs like GPT-4 have captured the public imagination with their verbal pyrotechnics, their propensity to hallucinate leaves them unsuited for the life-or-death realm of diagnostic medicine.
“We’re not trying to build a generalist AI,” Shah explains. “We’re building a specialist.”
That specialization seems to be paying dividends. In preliminary tests, Hippocratic AI’s model outperformed GPT-4 on over 100 medical certification exams — a crucial benchmark in establishing its healthcare bona fides. The company has assembled a team of physicians, nurses, and medical researchers to painstakingly train and validate its AI, aiming to create a system that can reliably handle low-risk, nondiagnostic tasks without putting patients in jeopardy.
“The No. 1 thing we care about is safety,” Shah stresses repeatedly. It’s a concern born partly from personal experience — Shah’s pivot to healthcare entrepreneurship came after his own brush with mortality at age 37 when chest pains during a run led to an ER visit the day after Google acquired his previous company.
That scare, which thankfully wasn’t a heart attack as he initially feared, sparked a personal wellness journey and a professional epiphany. “If I’m going to do something, I’m going to do it around health,” Shah recalls thinking. “I believe that you don’t find your mission in life, it finds you. I think circumstances drive it, and that was really the case.”
The Road to Hippocratic AI
Shah’s path to becoming a healthcare visionary was anything but linear. He cut his teeth in the frothy days of the original dot-com boom. His first startup, Andale, provided cloud-based tools for eBay and Amazon sellers — a prescient bet on the future of e-commerce in 1999.
After Andale’s acquisition, Shah struck gold again with like.com, a visual search engine that let online shoppers hunt for clothing and accessories using images instead of text. It was the kind of technically impressive, consumer-friendly innovation that caught Google’s eye, leading to that life-changing acquisition in 2010.
As generative AI models like GPT began demonstrating unprecedented linguistic abilities, Shah sensed an even greater opportunity on the horizon. And he decided to throw himself fully into the generative AI healthcare revolution.
“I started writing AI code in undergrad — it would have been 1994, I was writing neural networking code … on supercomputers,” Shah recounts. “Our neural network had 32 nodes, and it might have had 300 parameters. Today our models are 300 billion parameters, just to give you a sense of the scale. The AI just didn’t really work before. And then I saw ChatGPT, and I thought, ‘Oh, it really works.'”
A Solution in Search of a Crisis
Shah’s timing may prove fortuitous. The American healthcare system is under immense strain, with an aging population driving up costs even as shortages of nurses and other providers compromise care. The World Health Organization estimates a projected global shortfall of 10 million healthcare workers by 2030.
It’s against this backdrop that Shah envisions AI taking on a vast array of routine tasks. AI dietitians are available around the clock, automated lactation consultants for new mothers and conversational billing assistants to demystify medical charges. His ultimate goal is what he calls an “infinite pilot” model, where AI handles massive volumes of reminders, check-ins, and other nondiagnostic interactions that can improve patient outcomes without further taxing human resources.
“What if we had 350 million Americans and there were 350 million healthcare resources to help them?” Shah muses. “Would we get better outcomes? Of course, we would.”
He insists Hippocratic AI is taking a methodical, safety-first approach. The company’s large language model is trained exclusively on vetted, evidence-based medical information, not the free-for-all internet data that powers many general-purpose AIs. Human clinicians provide ongoing feedback and validation, aiming to create a system that knows its limits and won’t stray into diagnostic territory better left to human doctors.
Shah argues that by focusing on tasks that are currently going undone due to staffing shortages, Hippocratic AI will augment rather than replace human providers.
The Entrepreneur as Optimist
If Shah seems untroubled by the myriad challenges facing his ambitious venture, it may be because he’s seen this movie before. He draws parallels between the current AI boom and two previous technological watersheds he witnessed firsthand: the rise of the internet and the mobile revolution.
“I have never seen technology that’s made this much of a jump in capabilities,” Shah says. “These breakthroughs are like forest fires. Normally, when you try to create a company, it’s so hard to find a space in the forest where you can get sunlight that the big trees haven’t taken up. But when one of these big megatrends comes, it burns all the big trees. And your ability to create a new company is so much faster and so much easier.”
It’s a vivid simile that captures both the creative and destructive potential of transformative technologies. And make no mistake — if Shah’s vision comes to fruition, the impact on healthcare will be seismic.
But for all his talk of disruption, Shah comes across less as a brash techno-utopian than a builder genuinely driven by a desire to improve people’s lives. He speaks with equal passion about AI architectures and the nuances of chronic disease management. It’s clear that his 2010 health scare didn’t just redirect his career; it fundamentally altered his worldview.
“I completely revamped my diet,” Shah says of the aftermath of his ER visit. He threw himself into studying endocrinology and the biological systems underlying health. That personal journey informs his conviction that AI-driven personalization and constant availability can lead to better health outcomes.
What’s Next?
As Hippocratic AI enters a crucial beta testing phase with healthcare providers, the coming months will reveal much about the viability of Shah’s vision. The company’s recent partnership with Nvidia to power low-latency “empathy inference” for its AI agents hints at the scope of its ambitions. Shah isn’t just aiming to automate rote tasks; he wants to create digital health assistants that can engage patients with something approaching human-level emotional intelligence.
“I think my second mission in life probably just found me,” Shah reflects on his journey from e-commerce wunderkind to healthcare AI pioneer. The question now is whether doctors, regulators, and patients are ready for the AI-powered future Munjal Shah is racing to build.