Hippocratic AI, a company specializing in safety-focused large language models (LLMs) for healthcare, has received its first patent for the innovative Polaris architecture. This system, based on a “constellation” approach, employs a primary LLM alongside multiple support models to enhance accuracy and safety in healthcare applications. Initially, single LLM prototypes achieved only 80% accuracy in clinical conversations, but with Polaris, the company’s system surpassed human benchmarks, reaching 99.02% accuracy in its Polaris 2.0 version. The patented architecture includes specialized engines like the Overdose Engine, which detects potential medication toxicity, and the Medication Reconciliation Engine, which ensures accurate drug identification, among others. These components enable applications such as patient intake, pre-operative guidance, chronic care management, and insurance support, reducing healthcare costs and improving operational efficiency.
The Polaris system reflects Hippocratic AI’s mission to create a safe and impactful healthcare LLM. CEO Munjal Shah emphasized that the architecture exemplifies the company’s “do no harm” principle, enabling robust, empathetic, and clinically sound interactions. Founded by a multidisciplinary team of healthcare and AI experts, the company has raised $137 million in funding, backed by leading investors like General Catalyst and NVIDIA’s NVentures. Beyond the Polaris system, Hippocratic AI is addressing workforce shortages and enhancing care delivery by automating routine tasks, freeing healthcare professionals to focus on direct patient care. With its patented innovations, Hippocratic AI aims to transform global healthcare accessibility and outcomes.