Hippocratic AI powers safe, conversational voice agents for healthcare with Google Cloud

*Featured in this video:* Vishal Parikh, Co-Founder and Chief Product Officer of Hippocratic AI *About the company:* Hippocratic AI builds and deploys safe conversational voice AI agents that address healthcare's biggest problems with a mission to transform patient outcomes through healthcare abundance. Built on Polaris, the company's patented, safety tested constellation model, Hippocratic AI's agents have completed more than 270 million patient voice interactions across more than 300 live clinical use cases, with zero safety issues in production. Hippocratic AI has received a total of $444 million in funding and is backed by leading investors, including Andreessen Horowitz, General Catalyst, Kleiner Perkins, Avenir, NVIDIA's NVentures, Premji Invest, SV Angel, Google's CapitalG, and numerous health systems. Hippocratic AI stands behind its safety promise — its contracts allow the company to take responsibility for its agents. Founded on the principle "do no harm," the company's agents do not diagnose or prescribe. *Executive summary:* Hippocratic AI’s Polaris is a novel constellation architecture of more than 30 large language models (LLMs) optimized for healthcare settings, powered by Google’s AI Hypercomputer. Using Google Kubernetes Engine (GKE) and Dynamic Workload Scheduler, Hippocratic AI dynamically scales to meet the fluctuating demands of clinical settings while achieving natural, voice-based interactions with a time-to-first-token under 500ms and an average turn-to-turn latency of just 1.5 seconds. This architecture supports running over five trillion parameters for every exchange of dialogue and processes real-time AI workloads without manual oversight or paying for idle capacity — all while meeting the rigorous compliance standards required to operate over a thousand healthcare agents at scale. In addition, Google’s open Gemma models help Hippocratic AI to transform its high bar for clinical care into an operational reality, orchestrating millions of evaluations to maintain the highest levels of safety, clinical reasoning, and conversational accuracy. *Challenge:* Polaris is the world’s first constellation of over 30 specialized clinical models capable of safely performing patient-facing non-diagnostic tasks, from post-visit follow-ups to routine check-ins and medication adherence support. However, Hippocratic AI needed a flexible, open AI infrastructure capable of running trillions of parameters — on demand and in real time — without the high costs of idle capacity or the operational burdens of traditional setups. *Solution:* Partnering with Google Cloud, Hippocratic AI built Polaris on AI Hypercomputer, providing a powerful, secure-by-design foundation with built-in trust and governance that helps maintain strong healthcare compliance, such as HIPAA, HITRUST, and SOC 2. Instead of managing complex infrastructure, GKE and Dynamic Workload Scheduler allow teams to scale resources automatically, both on a schedule and in response to daily demand, ensuring AI agents are always available without having to pay for idle capacity. Hippocratic AI also uses Google’s open Gemma models as an “internal critic” to validate its AI models, orchestrating millions of evaluations to maintain its high bar for safety and clinical accuracy. *Results:* With Google Cloud, Hippocratic AI is leading the industry on conversational naturalness, achieving a time-to-first-token under 500ms and an average turn-to-turn latency of just 1.5 seconds. These real-time experiences are already proving their ability to help build trust and rapport with patents, with agents receiving on average an 8.9 out of 10 satisfaction rating. *Key takeaways and highlights from our interview with Vishal Parikh, Co-Founder and Chief Product Officer of Hippocratic AI:* → “We're able to deliver patient care using well over 50 clinical skills and actually move the patient through their health journey, and we do this conversationally, not just in text, but in voice so the patient can have a natural, free-flowing conversation.” → “GKE has given us the ability to actually deliver a large-scale system with a time-to-first token under 500 milliseconds, and an average turn-to-turn, end-to-end latency of around 1.5 seconds, which gives you both that safety and that real-time conversational experience.” → “Google Cloud has allowed us to run our system with much lower maintenance and engineering overhead, but also maintain the security and real-time constraints that we need at a cost-optimal point.” *Google Cloud products used: AI Hypercomputer, Dynamic Workload Scheduler, Google Kubernetes Engine, Gemma* *Learn more:* → hippocraticai.com