Hippocratic AI, the company pioneering the first safety-focused Large Language Model (LLM) for healthcare, has announced the issuance of its first patent, US Patent 12,142,371, granted by the US Patent Office this month. This patent covers key innovations integrated into Polaris, the company’s safety-driven LLM built using a constellation architecture designed specifically for healthcare applications.
The Polaris system’s constellation architecture is crucial in ensuring the safety and accuracy of Hippocratic AI’s technology. Early prototypes of the system answered only 80% of clinical questions accurately. However, after developing the constellation architecture, the Polaris 1.0 system outperformed human accuracy, and the latest version, Polaris 2.0, achieves a 99.02% accuracy rate on critical benchmarks.
The core concept behind the constellation architecture is that multiple LLMs are used in tandem to verify and double-check responses, enhancing both the safety and reliability of the system. Polaris 1.0, which launched in Q1 2024, has set a new standard, with many other AI agents now adopting a similar approach.
“Our patent highlights Hippocratic AI’s innovative approach to ensuring healthcare safety through our unique architecture,” said Munjal Shah, co-founder and CEO of Hippocratic AI. “Our team’s ability to develop, patent, and deploy Polaris in under two years demonstrates their vision and expertise, which are essential to our mission of ‘do no harm.’”
The patent covers critical elements of the Polaris system, including its primary and support models:
Primary Model:
- Conversation Interface: A custom-trained LLM designed for healthcare, ensuring empathetic and non-judgmental conversations while utilizing clinical techniques like motivational interviewing to enhance patient compliance.
Support Models:
- Overdose Engine: Monitors the conversation to detect potential medication overdoses or toxicity risks.
- Condition-Specific OTC Engine: Ensures patients are not advised to take over-the-counter medications that could be contraindicated based on their health conditions.
- Medication Reconciliation Engine: Verifies and corrects potential medication name errors, ensuring patients take the correct drug in the proper form.
In addition to these models, there are 16 other engines designed to enhance safety across various healthcare tasks.
The patent also includes applications for several key healthcare functions, such as:
- Preventative Screenings: Engaging patients for data collection on healthcare effectiveness.
- Intake and Scheduling Tasks: Assisting with appointment scheduling and health risk assessments.
- Pre-Operative and Discharge Tasks: Helping patients prepare for surgeries and manage post-operative care.
- Chronic Care Management: Reviewing medical records and providing self-management advice.
These applications aim to alleviate healthcare staff shortages, reduce costs, and improve patient engagement, ultimately enhancing healthcare delivery and patient satisfaction.
About Hippocratic AI
Hippocratic AI’s mission is to develop the first safety-focused LLM for healthcare, with the goal of improving healthcare accessibility and outcomes globally. The company was co-founded by Munjal Shah, along with healthcare professionals and AI researchers from prestigious institutions like El Camino Health, Johns Hopkins, and Stanford. Backed by $137 million in funding from leading investors including General Catalyst and Andreessen Horowitz, Hippocratic AI is driving innovation to revolutionize healthcare with advanced AI technology.
