Polaris 5.0 is here: Built on 200m+ patient interactions, outperforms every major model.

Book A Meeting
Login
Life Sciences

LLM Research

LLM Research

Our LLM research teams investigate the safety, clinical accuracy, and patient impact of healthcare AI models, so that artificial intelligence improves access to care as it becomes increasingly capable.

You are exceptional at software.
Now build the models behind it

Apply to Join HAI
paperscat
Hero banner with a blue gradient background title polaris a safety focused llm constellation architecture for healthcare on the left and three small document previews on the right featured badge top left
Safety
Multilingual

Polaris: A Safety-focused LLM Constellation Architecture for Healthcare

Lorem ipsum dolor sit amet, consectetur adipiscing elit. Ut elit tellus, luctus nec ullamcorper mattis, pulvinar dapibus leo.Lorem ipsum dolor sit amet, consectetur adipiscing elit. Ut elit tellus, luctus nec ullamcorper mattis, pulvinar dapibus leo.Lorem ipsum dolor sit amet, consectetur adipiscing elit. Ut elit tellus, luctus nec ullamcorper mattis, pulvinar dapibus leo.

Read More

Flowchart illustrating how a patients speech is transcribed by polaris and routed to a primary agent and a team of support agents
Safety
Multilingual

Polaris: A Safety-focused LLM Constellation Architecture for Healthcare

Lorem ipsum dolor sit amet, consectetur adipiscing elit. Ut elit tellus, luctus nec ullamcorper mattis, pulvinar dapibus leo.Lorem ipsum dolor sit amet, consectetur adipiscing elit. Ut elit tellus, luctus nec ullamcorper mattis, pulvinar dapibus leo.Lorem ipsum dolor sit amet, consectetur adipiscing elit. Ut elit tellus, luctus nec ullamcorper mattis, pulvinar dapibus leo.

Read More

Table of outcome measures with columns n mean sd 95 ci lowerupper rows include course satisfaction knowledge after courses content expected content relevance scenarios helpful outcomes met conflict disclosed ability to integrate learning commitment to apply notes that all measures use 5 point likert scales and n varies due to missing responses
Safety
Deployment

Transforming Health Care AI Education Through Micro-learning: A Novel Partnership Model for Nursing Workforce Development

Lorem ipsum dolor sit amet, consectetur adipiscing elit. Ut elit tellus, luctus nec ullamcorper mattis, pulvinar dapibus leo.Lorem ipsum dolor sit amet, consectetur adipiscing elit. Ut elit tellus, luctus nec ullamcorper mattis, pulvinar dapibus leo.Lorem ipsum dolor sit amet, consectetur adipiscing elit. Ut elit tellus, luctus nec ullamcorper mattis, pulvinar dapibus leo.

Read More

Scatter plot of ai models by latency ms and heart elo score with target latency and average human elo reference lines labeled points include gpt 4 chat claude gemini llama etc
Safety
Benchmarks

Perfecting Human–AI Interaction at Clinical Scale Turning Production Signals into Safer, More Human Conversations

Lorem ipsum dolor sit amet, consectetur adipiscing elit. Ut elit tellus, luctus nec ullamcorper mattis, pulvinar dapibus leo.Lorem ipsum dolor sit amet, consectetur adipiscing elit. Ut elit tellus, luctus nec ullamcorper mattis, pulvinar dapibus leo.Lorem ipsum dolor sit amet, consectetur adipiscing elit. Ut elit tellus, luctus nec ullamcorper mattis, pulvinar dapibus leo.

Read More

Venn diagram centered on the word heart showing five overlapping principles empathy attunement resonance human alignment and task following with bullet traits around each circle
Benchmarks
Safety

Perfecting Human–AI Interaction at Clinical Scale Turning Production Signals into Safer, More Human Conversations

Lorem ipsum dolor sit amet, consectetur adipiscing elit. Ut elit tellus, luctus nec ullamcorper mattis, pulvinar dapibus leo.Lorem ipsum dolor sit amet, consectetur adipiscing elit. Ut elit tellus, luctus nec ullamcorper mattis, pulvinar dapibus leo.Lorem ipsum dolor sit amet, consectetur adipiscing elit. Ut elit tellus, luctus nec ullamcorper mattis, pulvinar dapibus leo.

Read More

Process flowchart of a multi tier clinical review and update cycle from pre implementation to resolution in a structured diagram
Deployment
Safety

Real-World Evaluation of Large Language Models in Healthcare (RWE-LLM): A New Realm of AI Safety & Validation

Lorem ipsum dolor sit amet, consectetur adipiscing elit. Ut elit tellus, luctus nec ullamcorper mattis, pulvinar dapibus leo.Lorem ipsum dolor sit amet, consectetur adipiscing elit. Ut elit tellus, luctus nec ullamcorper mattis, pulvinar dapibus leo.Lorem ipsum dolor sit amet, consectetur adipiscing elit. Ut elit tellus, luctus nec ullamcorper mattis, pulvinar dapibus leo.

Read More

Line chart of em vs tool number 310 for six models claude 4 gemini 2 5 pro o4 mini deepseek qwen3 235b a22b kimi k2 em generally decreases as tool number increases
Multilingual
Safety

TRAJECT-Bench: A Trajectory-Aware Benchmark for Evaluating Agentic Tool Use

Lorem ipsum dolor sit amet, consectetur adipiscing elit. Ut elit tellus, luctus nec ullamcorper mattis, pulvinar dapibus leo.Lorem ipsum dolor sit amet, consectetur adipiscing elit. Ut elit tellus, luctus nec ullamcorper mattis, pulvinar dapibus leo.Lorem ipsum dolor sit amet, consectetur adipiscing elit. Ut elit tellus, luctus nec ullamcorper mattis, pulvinar dapibus leo.

Read More

Load More