Professor-led research collaboration

Turn your clinical question into a research instrument

We invite professors to lead AI research with us. The work begins with a question grounded in clinical practice. Under the professor's direction, we turn it into a research tool and the AI needed to test it.

Research questions arise in the course of care. A professor with deep experience in the work knows when to suspect a particular finding, which records to read together, and when judgment should be withheld. That knowledge is often at work in the clinic before it appears in a paper or guideline.

This is not a request to evaluate a finished product. The professor and research team lead the work: they set the question and interpret the findings. AetherHeal turns the question into a tool that can be tested and takes responsibility for the software, data processing, and research environment.

We are currently inviting professors to explore research in this form. We do not claim an active collaboration, IRB approval, data access, established clinical evidence, or regulatory status. A conversation with or participation by an individual researcher does not represent an agreement, endorsement, or recommendation by that person's institution. We name an institution only after an agreement is in place and the permitted scope of disclosure has been confirmed.

Why now

Clinical knowledge is no less valuable. What has changed is the cost structure of turning that knowledge into software that can be tested.

Until recently, turning a clinical question into software meant documenting requirements, finding a separate development team, establishing repositories and access controls, designing data schemas and annotation interfaces, fixing an execution environment, and managing errors and changes after deployment. Recruitment, outsourcing contracts, server procurement, security review, and maintenance could all precede the first test of the research question.

Generative AI and reusable development tools have reduced the work required to build a small research tool. An interface, data processing, annotation rules, and an evaluation environment can be held within one defined scope. The version used in a study can also be fixed and rerun under the same conditions. The investigator sets the clinical criteria and exceptions, decides what counts as input and when judgment should be withheld, and determines which results are meaningful.

AetherHeal is responsible for development and operation of the research environment. With the professor, we document the hypothesis, protocol, and rules governing how the tool behaves. We configure the environment within the approved study scope.

What we build together

We do not begin with AetherHeal's product catalogue. We begin with the research question, then define the inputs, unit of judgment, abstention criteria, and how the results will be interpreted before building the tool the question requires.

Review support for a defined procedure
Within investigator-defined inputs and findings, a tool could organize images, indicate locations and supporting evidence for review, and record the conditions under which interpretation should be withheld.
Follow-up measures for a defined cohort
From an approved cohort dataset, a tool could collect the measures specified in the protocol and identify missing fields and the next follow-up point.
A handoff structure for one specialty
A tool could structure the problems, evidence, incomplete work, and ownership that must carry through a handoff according to that specialty's workflow.

Research AI does not necessarily require a new model trained from scratch. Depending on the question, we select only the components needed: search and classification, annotation assistance, a workflow combining rules and models, or separate training. The choice follows the research question, the data, and reproducibility—not whichever feature is in fashion.

Bespoke development can be included in the research-support scope. That scope is limited to the question, approved data, users, functions, evaluation measures, fixed version, and exit conditions recorded in the protocol and agreement. It may include the interfaces, pipelines, annotation, analytical support, and deployment required for the research tool. Integration with a hospital's systems of record (booking, EMR, or CRM), hospital-wide release, service levels for clinical operations, and continuous maintenance are not treated as part of the same scope. Moving a research result into clinical operations requires separate security, regulatory, transition, and contractual review.

Who contributes what

The research team sets the clinical question and makes the research judgments. AetherHeal provides the engineering needed to turn those judgments into an executable tool. Neither party exercises the other's authority.

Responsibilities of the research team and AetherHeal by area
AreaProfessor and research institutionAetherHeal
Question and study designDefines the clinical question, hypothesis, intended use, inclusion and exclusion criteria, outcome measures, comparators, and clinical interpretation.Translates the question into implementable requirements, reviews technical fit and risk boundaries, and prepares the technical documentation required by the protocol.
DataSecures lawful authority to access the data and the basis for research use, and defines the included scope and institutional review process.Sets up the data pipeline, data dictionary, validation rules, and processing records within the approved scope. AetherHeal does not grant data-access authority on the institution's behalf.
Annotation and evaluationDefines the annotation criteria, clinical expert reviewers, adjudication process, and criteria for interpreting the results.Sets up the infrastructure for annotation interfaces, reviewer assignment, version control, disagreement tracking, and the history of evaluation outputs.
EngineeringExplains the research users' sequence of work and its exceptions, and reviews whether the tool remains within the clinical question.Takes responsibility for research interfaces and services, approved system connections, error handling, testing, deployment, and technical operation.
Reproducibility and changeApproves protocol changes and analytical decisions, and judges how a change affects interpretation of the findings.Provides an execution environment with fixed code, dependencies, data schemas, and configuration; preserves version and change records; and supports reruns under the same conditions.

How the data-governance terms are set

Before work begins, the parties define control of source data, who may access it, responsibility for de-identification, who leads institutional review, and what happens when the study ends. The terms below are our default proposal; final terms follow institutional policy, review decisions, and the research collaboration agreement.

Data-handling questions, default proposals, and documents to agree before work begins
QuestionDefault proposalAgree before starting
Where will the data reside?Our default proposal is to keep source data within the research institution's controlled environment. If external transfer is required, only the approved minimum dataset moves, under a separate agreement and the institution's transfer procedure.Data-flow diagram, storage locations, processors, and any cross-border transfer
Who may access it?Only people named, or assigned a named role, in the protocol and agreement receive least-privilege access. Before work begins, the parties define the purpose of access, how access is granted and withdrawn, and who is responsible for reviewing the records.Access list, role-based permissions, access records, and incident-response procedure
How will the data be de-identified?The research institution, or a party it authorizes, removes direct identifiers or pseudonymizes the data. The institution retains the re-identification key. Identification risks that may remain in dates, free text, audio, or images are reviewed separately.De-identification standard, custodian of the re-identification key, residual-risk review, and destruction method
Who submits for IRB review?For work involving human-participant data or medical information, our default proposal is for the principal investigator at the research institution to submit to that institution's IRB. AetherHeal provides the study plan, tool description, data-flow materials, and risk-management materials. The institution's own process also determines whether the work is exempt from review.Single-site or multisite structure, principal investigator, sponsor role, IRB costs, and responsibility for reporting changes
Can the data be reused for another purpose?Our default proposal is not to reuse research data or pseudonymized derivative data for model training or another study outside the agreed research purpose. Any reuse requires a separately defined institutional review and contractual scope.Permitted reuse of derivative data, models, and evaluation outputs, and the approval procedure
What happens when the study ends?Copies without a basis for retention are destroyed and access rights are revoked at the time specified in the agreement. Outputs needed to verify reproducibility require separate approval of the retention purpose, period, and access scope.Retention period, confirmation of destruction, and the retained scope of derivative data and the execution environment

Who owns the results, and how is authorship determined?

Technical support does not alter academic leadership of the research or the principle that findings may be disclosed. The collaboration agreement sets the allocation of specific rights and review periods before the study begins.

Academic leadership

The professor and research team retain academic leadership of the research question, study design, clinical interpretation, and manuscript. We propose terms under which neither the findings nor the manuscript becomes the exclusive asset of one party. Participating researchers have the right to analyze and present the findings within the agreed research purpose.

Data and intellectual property

The research institution controls the source data, and each participant retains the intellectual property it held before the study. The research collaboration agreement separately defines rights to use the research tool and source code, inventions arising during the study, generalizable foundation modules, and any reuse of pseudonymized derivative data, according to contribution and institutional policy.

Publication of negative findings

We propose terms that do not prevent presentation or submission when findings are negative or not statistically conclusive. AetherHeal will not request changes to a manuscript, delay disclosure, or seek withdrawal from publication because of the direction of the findings.

Authorship criteria

Authorship and author order follow the target journal's criteria and actual scholarly contribution. Roles are documented before the study begins and updated to reflect actual contribution before manuscript submission. Providing a tool, funding, or technical support alone does not establish authorship.

Review before submission

Each institution may review a manuscript before submission for factual accuracy, personal information, confidential information, and patentability. That review does not confer authority to change the interpretation of findings or refuse disclosure. The agreement sets the review period and any limited delay required for a patent filing in advance.

Where to begin

We are currently inviting professors and research teams to explore AI research with us. This does not mean that a study is under way with any institution, or that IRB approval, data access, clinical evidence, or regulatory review has been secured or completed.

  1. 01

    A one-page research question

    Set out the clinical question, the workflow currently performed by hand, the types of data that may be available and their access status, and the output the tool would need to produce.

  2. 02

    Scope and feasibility review

    Review whether the question can be implemented as a research tool, whether the required data and evaluation conditions can be established, and whether the scope is distinct from hospital-wide operational adoption.

  3. 03

    Protocol and agreement

    Document the roles, measures, data responsibilities, bespoke-development scope, costs, rights in the results, authorship criteria, and conditions for ending the study.

  4. 04

    Institutional review and environment preparation

    Access to data and setup of the research environment begin only after the required IRB and data reviews, contracts, and security review are complete.

Prepared by
AetherHeal
Document status
Research collaboration proposal

Send us a one-page research question, and we will review with you whether it can become an approvable study scope and an implementable research tool. Send enquiries to drjeeju@aetherheal.com.