en.Wedoany.com Reported - Hippocratic AI has announced the launch of a six-month LLM model training residency program, designed to enable experienced software engineers to transition into cutting-edge large language model (LLM) training while on the job, focusing on specific challenges in the voice AI domain. The program is open to both existing Hippocratic AI engineers and new hires, offering opportunities to develop advanced model training skills in the workplace.
Hippocratic AI is a generative AI company that has built a clinically validated, patient-facing voice AI system. Vishal Parikh, Chief Product Officer and co-founder, stated that the program aims to attract top engineers and apply their talents in system building and delivery to voice AI for improving patient care. Resident engineers will collaborate with the company's model organization, leveraging its infrastructure to carry out their work.
The residency program is not a lecture series; the work conducted during the residency has the potential to influence the development of the next-generation Polaris model and directly contribute to patient outcomes, with results potentially appearing in white papers and industry publications. Resident engineers will participate in model fine-tuning, run experiments on Hippocratic AI's NVIDIA H200 and B300 GPU clusters, and complete a capstone project, while engaging in one-on-one collaboration with senior machine learning engineers behind Polaris (the company's proprietary model cluster LLM). The program is extracurricular, requiring approximately 10 hours per week, alongside the resident engineer's primary role.
Hippocratic AI's engineering team backgrounds include institutions such as Stanford University, MIT, Carnegie Mellon University (CMU), UC Berkeley, and the Indian Institute of Technology (IIT). The company states that frontier foundation models are not yet sufficient for clinical voice AI, with accuracy rates in the medical domain falling far short of the required 99.9% standard. To address this, the company developed the Polaris cluster architecture, pairing a primary conversational model with specialized support models for medications, labs, dosages, social determinants, privacy, compliance, and safety. These specialized models check the primary model's handling of each clinical statement in real-time and deterministically. Polaris also drives a low-latency, empathetic, interruption-aware clinical-speed voice stack.
Subhabrata Mukherjee, co-founder and Chief Scientific Officer, stated that the engineering team builds voice AI from the model layer up on hardware, aiming to solve unresolved issues such as empathy, conversation quality, sub-second latency, and 99.9% accuracy. Resident engineers will directly contribute to this work.
Applications for the next round of the LLM model training residency program are open to Hippocratic AI employees and can be submitted through the company's careers page. This is the company's second residency program, following the Agent DE residency program for early-career engineers.
Hippocratic AI is a generative AI company dedicated to addressing the global shortage of clinicians, having built the patented safety cluster architecture Polaris, enabling non-diagnostic AI agents to support patients across multiple use cases through natural conversation. The company claims to have been tested by over 7,500 clinicians, completed more than 200 million patient interactions, maintained a 99.9% clinical safety record, and validated over 1,000 use cases across 60+ enterprise partners. Polaris 5.0 is a cluster with 5 trillion parameters, driven by a 700-billion-parameter core model, allowing the company to add new clinical skills and upgrade capabilities. According to the company, across more than 50 benchmarks, Polaris 5.0 surpasses leading deployable voice models from OpenAI, Anthropic, and Google on clinical, regulatory, and conversational tasks, while operating at speeds that frontier reasoning models (such as GPT 5.4 Pro, Claude Opus 4.7, Gemini 3.1 Pro) cannot match.
This article is compiled by Wedoany. All AI citations must indicate the source as "Wedoany". If there is any infringement or other issues, please notify us promptly, and we will modify or delete it accordingly. Email: news@wedoany.com
