Zack van Allen.

Applied AI · Primary care · Medical education

Zack van AllenPhD, MBA

I design and evaluate secure, human-centred AI for primary care and medical education, and study what it takes for it to be adopted well.

Senior Research Associate, University of Ottawa Department of Family Medicine and Bruyère Health Research Institute. My work brings together peer-reviewed research, implementation science and working prototypes, from clinician-led priority-setting and LLM evaluation to evidence-synthesis pipelines and AI governance.

2026–27 AMS Healthcare Fellow in Compassion & AI
Download CV Get in touch
Portrait of Zack van Allen
Peer-reviewed publications
28
First-author papers
9
Citations (Google Scholar)
1,800+
Years in health research
10+

Current focus · 2026–2027

Using secure, open-weight AI to protect compassionate care in family medicine

AMS Healthcare Fellowship Compassion & AI Principal investigator

I was named one of twelve fellows in AMS Healthcare's seventh cohort of Compassion and AI Fellows. The fellowship supports a year of protected time, hosted by the University of Ottawa Department of Family Medicine.

The project asks whether open-weight language models running on secure, local infrastructure can take on administrative work in academic family medicine, and whether doing so gives back the time and attention that compassionate care depends on.

AMS Healthcare announcement, September 2026

News

Recent updates

  • Sep 2026

    COMPASS I preprint posted on near-term AI capabilities and project ideas for primary care, from a virtual nominal group study with health care AI experts. Read the preprint

  • Sep 2026

    Named a 2026–27 AMS Healthcare Fellow in Compassion and AI, one of twelve fellows in the program's seventh cohort. Announcement

  • Jul 2026

    Viewpoint published in JMIR AI: how retrieval-augmented AI could turn static research outputs into living evidence for integrated knowledge translation. Read the paper

  • Jul 2026

    Protocol published in BMJ Open for a mixed-methods implementation study of Adaptive Mentorship Networks in Canadian primary care. Read the protocol

  • Jun 2026

    Co-organized Hackers & Healers, the inaugural AI-in-health-care co-design hackathon at Invest Ottawa: 150 participants in 44 clinician–developer teams turning clinician-identified needs into prototypes.

  • Apr–May 2026

    Workshop and talks at ICAM 2026 (Ottawa), the Ottawa Applied Medical AI Summit, and The Ottawa Hospital's Family Medicine Wellness Workshop. See talks

Research

Four connected lines of work

Applied AI projects that start from real clinical and educational problems, are built with the people who will use them, and are evaluated before anyone calls them solutions.

01

Secure, compassionate AI in primary care

Finding out what clinicians actually need, building privacy-preserving tools against those needs, and helping organizations adopt them responsibly.

COMPASS: from clinician needs to prototypes

Co-lead · ongoing

A discovery-to-prototype pipeline that starts from what clinicians want built rather than from a technology looking for a use case.

  1. 1 · DiscoverVirtual nominal group study with health care AI experts to rank near-term capabilities and project ideas.
  2. 2 · BuildHackers & Healers hackathon (June 2026): 150 participants, 44 clinician–developer teams.
  3. 3 · TestOnboarding clinics to test prototypes in real clinical settings (in progress).
COMPASS I preprint

Referral-triage application

Pre-production · under validation

A deterministic Python application that turns a clinic's manual intake, triage and routing workflow into explicit, reviewable logic, co-designed with clinicians and frontline triage staff.

Validation to date: 95% adjusted agreement with clinical triage decisions across 40 mock cases, using a standardized evaluation rubric.

Departmental AI adoption and governance

Lead author

An AI adoption plan for a multi-portfolio university department, built on a COM-B staff survey and aligned to the university's five-pillar AI roadmap: people, AI governance, data governance, technology and process. Endorsed in principle by senior leadership.

02

AI in medical education and assessment

Testing whether language models can support feedback, admissions and mentorship in medical training, and where human expertise still matters most.

LLM feedback on resident scholarly projects

Preprint

A rubric-aligned feedback assistant built on open-weight LLMs, tested in a blinded evaluation of 240 feedback reports: 120 LLM-generated and 120 expert-written.

Expert feedback was rated higher overall, and the gap varied by project stage and type, pointing to where expert review remains most important.

medRxiv preprint

AI-assisted admissions scoring

Evaluation

Evaluated an AI-assisted scoring pipeline for roughly 2,200 autobiographical sketches per admissions cycle, with bilingual support and equity auditing built into the evaluation framework.

Targets: ≥0.70 kappa and ≥0.80 AUC, with ~550 faculty hours a year potentially reclaimed.

Adaptive Mentorship Networks

Protocol published

A mixed-methods implementation science study of mentorship networks designed to strengthen primary care in Canada.

BMJ Open protocol
03

AI for evidence synthesis and qualitative methods

Human-in-the-loop workflows that make evidence faster to find, check and use, with provenance built in from the start.

Modernizing the Rourke Baby Record

Project design & oversight

An AI-assisted workflow to search, screen, verify and summarize pediatric evidence for the Rourke Baby Record, an evidence-based guide for well-baby and child visits in Canada, aiming to shorten update cycles from months to days while keeping every recommendation source-linked and expert-reviewed.

AI for integrated knowledge translation

JMIR AI 2026

A viewpoint arguing that retrieval-augmented LLMs, run as governed infrastructure with provenance, privacy protections and equity by design, can add a traceable conversational layer over research outputs for decision-makers.

Read the viewpoint

QualFrame and AI-directed content analysis

Research prototype

A developer-operated research prototype for human-supervised qualitative analysis, with provenance, versioned methods, QA checks, review queues and traceable exports, plus a long-form teaching guide to the method.

QualFrame on GitHub Methods guide

Universal Medical Ingestion Engine

Competition entry

Co-developed an entry for the Kaggle MedGemma Impact Challenge that turns complex medical documents into structured outputs for clinical and operational use.

Kaggle write-up
04

Health behaviour, aging and behavioural science

The foundation of my research training: modelling how health behaviours cluster and change, and what that means for interventions.

Machine learning with longitudinal aging data

CLSA

Prospective classification of functional dependence, pre-stroke physical activity and functional recovery, and pathways from pain to dependence in arthritis, using the Canadian Longitudinal Study on Aging.

Multiple health behaviours

Networks & clustering

How health behaviours cluster in more than 40,000 Canadians, temporal networks of health and protective behaviours during COVID-19, and systematic reviews of interventions that target several behaviours at once.

Behaviour change in health care

Implementation

Barriers and enablers to diabetic retinopathy screening, hepatitis C care, vaccination and plasma donation, informed by behavioural theory and patient-oriented research.

Publications

Research outputs

28 peer-reviewed publications, including 9 as first author, plus recent preprints. Over 1,800 citations on Google Scholar.

Peer-reviewedViewpoint2026

From Static Outputs to Living Evidence: AI for Integrated Knowledge Translation in Canadian Health Research

JMIR AI · first author

Makes the case for retrieval-augmented language models as governed, auditable infrastructure that lets decision-makers query curated research outputs between meetings, with provenance and privacy built in.

Read the paper
Preprint2026

COMPASS I: Near-Term AI Capabilities and Project Ideas for Primary Care

JMIR Preprints · first author

Nine health care AI experts, five of them physicians, rated AI capabilities and project ideas for primary care. A digital medical office assistant and an inbox manager ranked highest, pointing to administrative relief as the near-term priority.

Read the preprint
Preprint2026

Can AI Match Human Experts? Evaluating LLM-Generated Feedback on Resident Scholarly Projects

medRxiv · first author

A blinded comparison of 120 LLM-generated and 120 expert-written feedback reports. Experts were rated higher overall, and the difference depended on project stage and type.

Read the preprint

All publications

2026

2025

2024

2023

2022

2021

2020

2019 and earlier

Also: government and scientific reports on COVID-19 behavioural science and military leadership, and doctoral and master's theses. Full list in the CV, with citation metrics on Google Scholar and ORCID.

Talks & teaching

Talks, workshops and teaching resources

  1. May 14, 2026

    AI, Administrative Burden and Physician Wellness in Family Practice

    Invited workshop, Family Medicine Wellness Workshop, The Ottawa Hospital Department of Family Practice, Ottawa

  2. May 12, 2026

    AI-Driven Automation of Autobiographical Sketch Grading in UGME Admissions

    Ottawa Applied Medical AI Summit, hosted by the Ottawa Medical AI Research Institute (OMARI) and CHEO Research Institute

  3. Apr 2026

    Ethically Integrating AI Into Deductive Qualitative Analysis in Medical Education

    Workshop co-lead, International Congress on Academic Medicine (ICAM 2026), Ottawa

  4. Jun 4, 2025

    Artificial Intelligence in Elderly Care: Looking to the Future

    Invited talk, 16th Annual Care of the Elderly Conference, McMaster University

  5. Apr 27, 2025

    Artificial Intelligence in Primary Care: Envisioning the Future Integration of AI in Family Medicine

    Invited talk, Department of Family Medicine Faculty Retreat

  6. Sep 24, 2024

    The Impact of Physical Activity Pre-Stroke and the Insights From Machine Learning on Functional Dependence

    Invited webinar, Centre of Excellence Webinar Series, Perley Health

  7. Mar 17, 2022

    Temporal Network Dynamics of Multiple Health Behaviours During the COVID-19 Pandemic

    Invited talk, Montreal Behavioural Medicine Centre

Funding

Fellowships and grants

Personal fellowships where I am the principal investigator, and team grants where I am AI lead or co-investigator and others hold the PI role.

$320,000
Personal fellowships and awards
3 competitive awards as principal investigator
$1.97M
Team grants
8 funded projects as AI lead or co-investigator

Personal fellowships

  • AMS Healthcare Fellowship in Compassion and Artificial Intelligence AMS Healthcare · 2026–2027 $75,000
  • Mitacs-Banting Discovery Postdoctoral Fellowship Physical activity engagement and the intention–action gap in older adults $140,000
  • Frederick Banting and Charles Best Canada Graduate Scholarship – Doctoral CIHR · Complexity science and network analysis for health behaviour change $105,000

Team grants

  • Partnering with patients to co-produce a national AI priority agenda for chronic pain CIHR Catalyst Grant: Partnering for Impact · 2026 $125,000
  • Optimizing pandemic preparedness: extending the iCARE study CIHR · public attitudes, intentions and behaviours $952,426
  • POPCORN reporting guidelines for modelling studies CIHR · noncommunicable disease modelling $320,536
  • Co-designing francophone AI health tools with patients and caregivers Association médicale universitaire de Montfort (AMUM) $186,120
  • AsthmaWISE: a multimodal conversational AI companion for asthma CHAMO $135,000
  • Modernizing the Rourke Baby Record: AI-enabled evidence synthesis TOHAMO $135,000
  • EmPOWERing primary care settings to use digital health interventions for chronic pain CIHR Catalyst Grant · 2025 $96,788
  • COMPASS: Consensus on Medical Priorities and AI Solutions in Primary Care The Ottawa Hospital seed fund $15,000

About

Experience and training

Trained in health psychology and behavioural science, with a PhD focused on machine learning and network models of health behaviour, and an MBA.

  1. 2024–present

    Senior Research Associate

    University of Ottawa Department of Family Medicine and Bruyère Health Research Institute
    • Lead applied AI work across both institutions, from clinician priority-setting to prototypes and evaluation.
    • Built a deterministic referral-triage application with clinicians (pre-production, under validation).
    • Evaluate LLM tools for medical education, including resident feedback and admissions scoring.
    • Authored the department's AI adoption plan and advise leadership on AI adoption.
    • Provide research-methods consulting and mentorship to faculty, physicians and research staff.
  2. 2023–2024

    Mitacs-Banting Discovery Postdoctoral Fellow

    University of Ottawa
    • Machine learning and longitudinal modelling of aging, stroke and functional dependence.
  3. 2018–2024

    Clinical Research Coordinator, then Clinical Research Associate

    Ottawa Hospital Research Institute
    • Led a multi-province patient-oriented project on diabetic retinopathy screening.
    • Behavioural science evidence syntheses, cluster and network analyses, and machine learning for healthy aging.
  4. 2015–2018

    Earlier roles

    University of Ottawa · Department of National Defence
    • Crisis resource management and stress-inoculation training modules for medical students.
    • Co-authored scientific reports on leadership profiles for senior Canadian Armed Forces positions.

Education

  • Postdoctoral Fellowship

    University of Ottawa · 2023–2024 · Mitacs-Banting Fellow

  • PhD, Psychology

    University of Ottawa · 2019–2023 · CIHR Doctoral Scholarship. Dissertation: Modelling co-occurring and co-varying health behaviours: Applications of machine learning and network psychometrics

  • MBA

    Carleton University · 2017–2019

  • MA, Psychology

    Carleton University · 2014–2016

  • BA, Psychology

    Carleton University · 2009–2014

Methods and tools

Python, R, SQL · LLM evaluation, retrieval-augmented generation, NLP · predictive modelling, multilevel and network models · systematic reviews and meta-analysis · qualitative and mixed methods · implementation science (COM-B, TDF)

Contact

Interested in collaborating?

I'm glad to hear about research partnerships, speaking and workshop invitations, and applied AI projects in health care and medical education.

vanallen.22@gmail.com Download CV