Skip to content
ushealthcloud

AI Solutions

Pre-trained agentic AI models for care, coverage and research

Start from models built for healthcare language and workflows, then adapt them to your data in a HIPAA-compliant environment, with the documentation and oversight the EU AI Act expects.

Model families

Agents that know the work, and when to stop

Three pre-trained families, each adapted to your workflow. Every run ends the same way: a person makes the decision.

We adapt and deploy agents for intake, documentation drafting and care-gap outreach on your data, and they hand decisions back to clinicians.

ush-agent · care deliveryExample · synthetic
  1. inputreferral_0193.pdf received · fax, 6 pages
  2. agentread document · extract diagnosis, service, payer
  3. toolehr.match_patient() · matched to MRN
  4. guardPHI stays inside the HIPAA-compliant zone
  5. agentdraft intake summary for the care team
  6. handoffclinician reviews and decides

Adaptation

Adapted to your data

Fine-tuning on de-identified data inside a HIPAA-compliant environment, evaluated against your own benchmarks before deployment.

  1. 01

    Base

    Pre-trained healthcare model

  2. 02

    Tune

    On de-identified data

  3. 03

    Evaluate

    Against your benchmarks

  4. 04

    Deploy

    With human review

Before go-live, every model passes

  • Accuracy on your own cases
  • Failure modes tested and documented
  • PHI leakage tests
  • Prompt-injection tests
  • Human oversight defined
  • Risks and limits documented
Works with
  • Clinical notes
  • Discharge summaries
  • Prior-auth criteria
  • Claims lines
  • Benefit documents
  • Trial protocols
  • Literature abstracts
  • Patient messages

How it works

A loop, not a launch

Models change, policies change and workflows change. Every change sends the model back through evaluation.

  1. 01

    Assess

    Choose the workflow and the model family, and define what good looks like with the people who will use it.

  2. 02

    Adapt

    Fine-tune or configure on de-identified data, connect tools, and set the actions an agent may and may not take.

  3. 03

    Validate

    Evaluate on your cases, including failure modes, and document risks, limits and oversight before go-live.

  4. 04

    Operate

    Monitor outputs and overrides, collect feedback and re-evaluate after every model or policy change.

FAQ

Questions we hear from compliance and data teams

  • Data useAre your models trained on our data?

    Only if you choose to fine-tune, and only under the terms of your agreement. Fine-tuning uses de-identified data by default.

  • OversightCan agents take actions in our systems?

    Only the actions you approve. Each agent has an explicit list of tools it may use, and high-impact actions require a person to confirm.

  • RegulationWhat does EU AI Act alignment mean in practice?

    Documented risk management, data governance, human oversight, transparency to users and ongoing monitoring, the practices the Act sets out for higher-risk AI systems, applied to how we build and deploy models.

  • EvaluationHow do you evaluate a model before deployment?

    Against task-specific test sets drawn from your workflow, including edge cases and known failure modes, with results documented and reviewed by your clinical or operational owners.

Next step

Bring us the data problem your compliance team worries about most.

We'll map where protected health information flows today, and how to model, automate and secure it.

What the first call gives you

  1. 1A map of where PHI flows today
  2. 2One workflow worth starting with
  3. 3The controls that workflow needs

A short scoping call, then a written proposal. Please don't send patient information.