Hippocratic AI: This Company Is Building A Safety-Focused LLM To Improve Healthcare

By Amit Chowdhry • Updated September 22, 2023

Hippocratic AI is a company that is building a safety-focused large language model (LLM) for the healthcare industry. Pulse 2.0 interviewed Hippocratic AI CEO and co-founder Munjal Shah to learn more.

Munjal Shah’s Background

Munjal Shah

Shah received a Bachelor’s degree in Computer Science from the University of San Diego and a Master’s in Computer Science from Stanford University, specializing in AI. Shah said:

“After my studies, I became a serial entrepreneur, startup advisor, and investor. Hippocratic AI is the fourth company I founded since beginning my entrepreneurial journey in 1999. The second company I founded, a machine learning computer vision company, gained rapid popularity and was acquired by Google in 2010. However, the day after the deal closed with Google, I had a significant health scare. While going for a run, I started experiencing chest pains and had to go to the ER.”

“What should have been the best day of my career actually ended up being one of my worst. I spent the entire day in the ER, knowing that my Dad had experienced something similar as he had his first heart attack in his 40s. I was 37. Ever since I changed my entire worldview when it came to my personal health and the broader health system as well. Since then, I’ve lost 30 pounds, worked out almost daily, and started taking classes in medicine and endocrinology. This motivated me to merge my newfound passion for healthcare with my entrepreneurial expertise.”

“Over the next decade, I spearheaded a few different companies, including a healthcare startup that leveraged AI to analyze health records, a machine learning company acquired by Google, and a marketplace management, research, and analytics company that eventually sold to Alibaba. Recognizing the potential of applications for generative AI in healthcare, I decided to create Hippocratic AI to be the first safety-focused large language model (LLM) for healthcare. Founded alongside a group of physicians, hospital administrators, Medicare professionals, and artificial intelligence researchers, we’re on a mission to dramatically improve healthcare access, equity, and outcomes through the use of generative AI.”

Solving Healthcare Industry Challenges

What are the primary challenges in the healthcare industry you’re solving? Shah noted: 

“Right now, we are facing a massive worldwide healthcare worker shortage. In fact, the World Health Organization predicts that there will be a shortfall of 15 million healthcare workers globally by 2030. Many people have spent years of their careers trying to fix healthcare access within the system, but we’re still nowhere close to where we need to be.”

“Now, we finally have the technology needed to be able to deliver affordable healthcare, especially to underserved communities and geographies. This is where I see Hippocratic AI coming in. We are finally able to leverage generative AI to massively increase healthcare access for millions of people around the world. With this technology, I believe we can close the healthcare gap and provide a level of care that’s never existed before – I like to call this ‘super-staffing.’ We envision Hippocratic AI working alongside healthcare workers of all kinds to amplify their quality of care and uplevel their work in engaging with patients. I firmly believe that Hippocratic AI is the best solution to solve this massive shortfall and the only way to truly bring healthcare to every home.”

Core Products

What are Hippocratic AI’s core products and features? Shah explained:

“Hippocratic AI will be used for non-diagnostic patient-facing applications. Luckily, there are thousands of healthcare LLM use cases beyond just diagnoses that our model will be able to support. This includes an explanation of benefits or billing, dieticians, genetic counselors, pre-op questions, delivering test results that are negative (e.g., we found nothing wrong), appointment reminders, care instructions and communications, risks assessments, and many others. Additionally, the model has been trained to understand the wide spectrum of healthcare jargon, terminology and dialect, such as abstract prescription names like ‘Albuterol.’”

“Through extensive training, Hippocratic AI’s LLM will revolutionize patient interactions. AI capabilities will be incorporated for voice and tone detection, enabling the model to comprehend emotions and tones that are not conveyed through text alone. This understanding will enhance the quality of care provided, allowing the LLM to detect things like fear or pain, amongst other common patient emotions. Further, it will allow for deeper patient connection and provide ongoing care that goes beyond the transactional level. Many times, patient satisfaction is hindered by being interrupted by a healthcare practitioner. On average, a doctor interrupts a patient after 11 seconds of their interaction. In contrast, the LLM has unlimited time to listen, comprehend, and address patient needs in a way that can be tailored to each individual patient.”

“Our goal is to work with healthcare practitioners of all kinds to help alleviate pressure on health staffing teams, decrease patient wait times, and enable staff to devote more attention to each patient they see. This is the main reason why we are focused on low-risk healthcare tasks that will ensure patient safety – while also improving healthcare access and overall outcomes.’


In May, Hippocratic AI announced a $50 million all-equity seed round led by General Catalyst and a16z. Shah pointed out:

“This has been the first investment into the company, and we are going to be using the funding to invest heavily in talent, computing, data, and partnerships. The money, time, and resources invested into our company will go directly towards furthering our mission of fundamentally increasing the supply and scalability of healthcare professionals to achieve a more proactive, affordable, and equitable system of care for all.”

Formation Of Building The LLM

Can you walk us through the process of building the LLM? Shah revealed:

“Throughout the entire development process, we focused on three core areas: certification, reinforcement learning with human feedback (RLHF), and bedside manner. Focusing on these three main parameters ensured that we were able to set ourselves up for success and give us the best possible chance to reach our goal of using the LLM to increase healthcare access, reduce cost, and close the healthcare skills gap that the pandemic created.”

“Certification: Research has shown that existing AI models can pass the USMLE (US Medical Licensing Exam). However, we knew that passing the USMLE is not enough to ensure a model is ready for the wide variety of healthcare roles that exist in care and payer settings. Therefore, we set out to build our own model that can test on not just one but 118 different healthcare certifications and roles. This was all with the goal in mind of building a viable product that would outperform existing state-of-the-art language models like ChatGPT and GPT4, and other commercially available models. We did this by training our model on a diverse set of certified, evidence-based content, unlike some of the other models on the market today. One of our biggest markers of success is that our model is proven to outperform GPT-4 on 110 of the 118 tests and certifications – 5% on 78 of the certifications, and 10% on 47 of their certifications.”

“RLHF with Healthcare professionals: One of the most important parts throughout the building process was utilizing RLHF and tapping into the human aspect when developing the LLM. Since the start of Hippocratic AI, we have engaged with many different healthcare professionals to help guide and train the LLM by rating thousands of responses. At the end of the day, we know that the best people to determine LLM readiness for deployment in the healthcare system are the experts who serve in that role in today’s system.”

“Bedside manner: In healthcare, it’s equally important to have accurate responses and a compassionate approach when speaking to patients. Bedside manner significantly impacts the patient’s emotional well-being and quality of outcomes. This applies to all healthcare professionals, not just doctors. Currently, there are no benchmarks for evaluating the bedside manner of a language model when interacting with patients. However, we will be releasing the first of many bedside manner benchmarks for the entire community to use. This includes: showing empathy, care and compassion, making a patient feel at ease, taking a personal interest in a patient’s life, and helping patients take control.”

“With Hippocratic AI, we’ve marked the beginning of our vision to use language models to massively increase healthcare access, reduce costs, and close the healthcare skills gap left behind by the global pandemic. LLMs are one of the best new ways to achieve this, but it has to be done in a safe way that’s specifically tuned for the healthcare industry.”

Future Company Goals

What are some of Hippocratic AI’s future company goals? Shah concluded:

“The healthcare industry needs its own AI platform, one that is focused on empowering the workforce, reducing burnout, and improving patient safety and experiences with the healthcare system. Ultimately, we see Hippocratic AI having a role in fundamentally increasing the supply and scalability of healthcare professionals. It’s our commitment to extend Hippocratic AI to roles and tasks only when the system is guaranteed safe and accurate by qualified healthcare professionals.”