Research · Open source · Frontier AI
We build the future of healthcare in the open.
Universal Prognostics is a research-driven company. We run live AI agents and fine-tuned models inside real healthcare operations — and we push the frontier of healthcare AI in public, with the people and institutions advancing the field.
Why we research
Intelligence, deployed for good.
The best healthcare AI should not sit in a paper or a private lab. We deploy it — in labs, clinics, and the operating systems that run them — and we open the foundational research so the whole field moves faster. We love working with people: clinicians, scientists, and engineers building toward the same goal.
Featured open-source project
AutoScientists
A decentralized team of AI agents for long-running computational scientific experimentation.
How Universal uses AutoScientists
Shared State
Champion p*
no champions yet
performance metric
Log
experiments
Forum
posts
Discussion
attention
RUNNINGExperiments
regularization
QUEUEDExperiments
Decentralized agents read a shared state, self-organize into teams around emerging research directions, run experiments in parallel, and reorganize when a direction stagnates.
Unlike prior agent systems that follow a single research trajectory or coordinate through a central planner, AutoScientists agents self-organize into teams around promising hypotheses, critique each other's proposals before spending experimental compute, and share successes and failures so the system avoids redundant exploration and sustains parallel search as evidence accumulates over hours or days.
Self-organizing teams
Agents cluster around promising hypotheses instead of following one fixed trajectory or a central planner.
Critique before compute
Proposals are challenged by peer agents before any experimental compute is spent.
Shared wins and failures
Successes and dead ends are broadcast across the swarm, so it avoids redundant exploration.
Sustained parallel search
Many lines of inquiry run at once and persist as evidence accumulates over hours or days.
Developed in collaboration with Harvard.
What we are building
Research that reaches patients.
We push the frontier in public — and deploy what works inside real clinics, labs, and hospital workspaces.
In-house healthcare LLM
We are fine-tuning our own large language model on UP operating and clinical workflow data — built for the tasks generic models cannot do well inside healthcare.
AI radiology platform
A candidate digital diagnostics layer for Universal X Cloud — connecting imaging intelligence to the same data context that runs our labs.
Local models at the edge
Private models deployed inside clinics, labs, and hospital workspaces — on-site intelligence that keeps sensitive data local while staying connected to the platform.
AutoScientists
Decentralized agent teams that self-organize around research directions — the same architecture we use to continuously improve screening and operations models.
Live in production
Real intelligence, running real operations — today.
These are not roadmap slides. UP runs autonomous agents and fine-tuned models across the live lab network right now.
01
Autonomous agents
Agents orchestrate supply and reagent forecasting, quality-control anomaly detection, scheduling, and deployment planning across the network.
02
Fine-tuned LLMs
Proprietary models fine-tuned on UP's operating and healthcare data for domain tasks generic models can't do well.
03
Workflow assistance
AI tools embedded in HIS/XCloud that assist lab techs, intake, reporting, and multilingual patient communication.
04
Screening & diagnostics models
AI-supported screening connected to diagnostic confirmation pathways — supporting clinicians, never replacing them.
Clinical and regulatory guardrail
Our clinical AI supports workflows and confirmation pathways. It does not replace clinicians, regulated diagnostic confirmation, pathologists, or laboratory quality systems.
Work with us
Researchers, clinicians, and builders — let's push the frontier together.
We love collaborating: open-source contributors, academic partners, and people who want their work to reach real patients in real markets. If you're advancing healthcare AI for good, we want to hear from you.
Start a research conversation