From Validated Models to Agentic AI for Healthcare Applications
Are we ready for agentic health assessments?
Are we ready for agentic health assessments?
Medical AI still runs mostly on static risk models built from curated risk factors. Agentic AI systems can read a patient's full history, reason about it, search for outside evidence, and call validated models as tools.
That opens a path to better risk assessment for cancer, diabetes, and cardiovascular disease. It also raises hard questions about calibration, provenance, privacy, and accountability, and about whether these systems work for Native Hawaiian, Filipino, and Pacific Islander patients, who are missing from most training data.
The workshop asks what a safe, fair, and effective framework for agentic risk assessment needs, and builds the Hawaiʻi collaborations to answer it.
From the research
“Clinical predictive models are frequently derived from rules that have existed for decades”
“AgentMD can automatically select and apply the relevant RiskCalcs tools given any patient description.”
“the research protocol compressed timelines from months to hours”
“Are we ready for agentic health assessments?”
From calling validated models to building new ones
Aggregation
Agents that run validated clinical calculators and risk models as tools.
Novel agents
Agents that build their own risk estimates from the full patient history.
Validation
Digital twins, patient similarity, and "patients like me" evidence.
Four invited keynotes, each moderated by a scientist from one of our industry partners. Speakers will be announced.
Friday, February 12
- 7:30 AM Registration, coffee, and poster setup
- 8:00 AM Welcome
- 8:15 AM Opening keynote: Language models as prediction engines for a health systemSpeaker to be announced
- 9:10 AM Lightning talks: 90 seconds per poster
- 9:25 AM Coffee and poster lounge
- 9:55 AM Keynote: Agents that call the validated risk modelsSpeaker to be announced
- 10:50 AM Lightning talks
- 11:05 AM Poster lounge
- 11:35 AM Lunch with sponsor tables, at the John A. Burns School of Medicine
- 12:50 PM Keynote: Agents that read the whole record and make their own estimateSpeaker to be announced
- 1:45 PM Lightning talks
- 2:00 PM Coffee and poster lounge
- 2:30 PM Next Wave Transformative Award finalists: three early-stage investigators10-minute presentations, each followed by 5 minutes of questions
- 3:15 PM Keynote: Are we ready for agentic health assessments? AI governance, ethics, and responsibilitySpeaker to be announced
- 4:10 PM Awards: poster awards and the $50,000 Next Wave Transformative Award
- 4:30 PM Closing remarks
- 4:40 PM Reception for all attendees, with the speakers
Times are Hawaiʻi Standard Time. The program is in the Sullivan Conference Center, except lunch, at the John A. Burns School of Medicine next door.
The $50,000 Next Wave Transformative Award
A single competitive $50,000 award for agentic AI that transforms health care in Hawaiʻi and the Pacific.
This is a one-time award for the workshop, separate from PACAID's NIH-funded pilot grants. It is open to NIH Early Stage Investigators at the University of Hawaiʻi. Three finalists present their proposals at the workshop, and the award is made on the day.
Letters of intent are due Friday, January 8, 2027. The request for applications opens December 7.
The road to February 12
Two ways to take the stage. Trainees bring a poster and a 90-second lightning talk. Early-stage investigators compete for the $50,000 Next Wave Transformative Award. Both tracks open December 7 and meet at the workshop.
AbstractsPosters and lightning talks
- Dec 7Call opens
- Jan 15Abstracts due
- Jan 22Acceptances
- Feb 5Lightning slide due
- Feb 12Talk and poster
Transformative Award$50,000 for early-stage investigators
- Dec 7Call opens
- Jan 8Letter of intent due
- Jan 153 finalists named
- Jan 29MEDAI Core consult
- Feb 5Proposal and slides
- Feb 12Present; award made
Friday, February 12, 2027Sullivan Conference Center, University of Hawaiʻi Cancer Center. Register by January 29.
Join us on February 12
Register
Free for UH students, faculty, and staff. A modest fee for other attendees; details soon.
Submit an abstract
Every accepted poster gets a 90-second lightning talk. Abstracts due January 15.
Apply for the Next Wave Transformative Award
$50,000 for early-stage investigators. Letters of intent due January 8.
Stay informed. Join the PACAID mailing list for workshop announcements and monthly PAC-AID Talks.
Sign upUH faculty, postdocs, and students in medicine, public health, computer science, engineering, and data science; clinicians and health system leaders in Hawaiʻi and the Pacific; and industry researchers building AI for health.
Hosted by PACAID, the Pacific Center for Artificial Intelligence and Data Science in Medicine (NIH/NIGMS COBRE P20GM161995), at the University of Hawaiʻi. Supported by the UH Office of the Vice President for Research and Innovation. Organized by Peter Sadowski and Yujia Qin of the PACAID MEDAI Core, with John Shepherd and Youping Deng.