Clinical innovation · Research and evaluation
I help clinical organizations improve how data is collected, reviewed, and used—through better workflows, purpose-built software, and targeted uses of AI. The result should be something their teams can run themselves. I also lead organizational research and evaluation from the research question and study design through publication and presentation.
Practice B
Research & evaluation for organizations
A product, program, or clinical method should not become persuasive faster than it becomes testable.
I work with clinical organizations, product teams, and university groups that need evidence for an important decision. I can help define the question, design the comparison, choose measures and analysis, and turn the result into figures, an argument, a publication, or a presentation.
Work can begin before data collection, while the design can still change. It can also begin later, with difficult results, a manuscript, a figure, or a talk that is not yet clear. The goal is evidence another researcher, reviewer, leader, or audience can inspect and use.
01
Define what the organization needs to learn, what should be compared, and which competing explanations the study must rule out.
Research question · comparison · study protocol
02
Choose measures, participants, controls, data structure, and analysis before the result is known—and build in checks for data quality and bias.
Study design · measurement plan · analysis plan
03
Analyze whether the finding holds across difficult cases, plausible alternatives, and sensitivity checks, then connect it back to the decision.
Analysis · figures · recommendation
04
Build the manuscript, technical report, figures, talk, or poster around what the evidence supports, then prepare the team for reviewers and live questions.
Manuscript · presentation · reviewer response
Why trust the work
My published work addresses each point: testing whether a measure supports the claims placed on it, committing the design before results are known, and turning technical findings into arguments that reviewers, leaders, and live audiences can challenge and use. The broader record includes five peer-reviewed articles, open datasets, university teaching in research methods and statistics, and international presentations.
Measurement validity
A first-author validity study tested whether a measure intended to quantify decision-making could be deliberately manipulated. Across 200 participants and two randomized experiments, it examined the practical risk of both false positives and false negatives. For an organization, that is the difference between collecting data and knowing whether a decision can defensibly rest on it.
Read the article →Research integrity · Cortex
For a Cortex Registered Report, the research question, measures, and analysis were accepted before results were known, and the data and analysis were published openly. That structure made an unexpected result useful: it clarified what the evidence could support without quietly changing the question after the fact.
Review the study →Scientific communication
Work from the same research program became Best Talk by a Trainee at the 2019 ISPGR World Congress. My broader presentation record also includes third place in the Noba + Psi Chi Student Video Award and finalist recognition in the APA Psychological Science Video Festival. Strong evidence still has to become a case that reviewers, leaders, and live audiences can follow and question.
See the full record →Practice A
Clinical systems & AI
Clinical innovation starts with one workflow that needs to work better—not with an organization-wide promise to “adopt AI.”
Clinical teams collect data while delivering care, turn observations into graphs and reports, supervise with limited time, and remain accountable to families, funders, and their own standards. Any technology introduced here has to respect those constraints.
A partnership starts with the people doing the work. Together, we identify where a purpose-built tool or AI could save staff time, reduce errors, or expand capacity, then test it against the current workflow before anyone commits to a wider rollout.
Start with the costly, unreliable, or repetitive part of the work—not with a preferred model or vendor. Map who does it, where it breaks down, and what a better result would look like.
Workflow map · priorities
Decide what data is needed, how work will move, what must stay private, where errors matter most, and which decisions remain with people.
Requirements · prototype plan · responsibilities
Compare the new approach with the current workflow using representative cases, agreed measures, staff feedback, and clear stop rules.
Pilot · results · recommendation
If the pilot works, put the training, monitoring, escalation, documentation, and ownership in place so the organization can run it without me.
Training · safeguards · handover
Concrete starting points
Make it easier to record what happened without pulling attention away from care. Design collection and review tools around the measures and decisions the clinical team actually uses.
Explore transcription, drafting, and review tools that reduce repetitive documentation and help supervisors see what needs attention—without outsourcing judgment.
Connect observations, graphs, review notes, and decisions so teams can see both the source data and how it was interpreted.
When existing products do not fit, design the smallest tool or integration that solves the problem, then test whether it merits wider use.
I reserve a small number of design partnerships for organizations ready to test a specific workflow and act on the result. I use the same approach in Formative Grapher and BehaviorStream: privacy, human review, error handling, and ownership are built in from the start. It also shaped the Executive AI curriculum for leaders deciding where to use AI, what to build or buy, and what not to automate.
Fit
How an engagement begins
What makes the work productive
The work affects care delivery, measurement, a product, a study, a program, or an operating workflow—not a generic desire to “use AI.”
The staff who use the process can help define the problem and test whether the change works.
Start with one workflow, study, comparison, or deliverable that is small enough to test and important enough to matter.
Someone inside the organization owns the decision and can act on the result.
First step
A short brief is enough: what is being decided, what exists now, what constraint matters, and who will use the result. I will tell you whether the work fits and what a sensible first engagement would be.
I also consider focused projects outside clinical care when they combine operations, technical implementation, and rigorous evaluation.