01
Problem definition
Map the user, intended use, decision being supported, failure consequences and the current non-AI workflow.
Applied AI and software development
Responsible technology for football medical workflows.
Clinical context first: define the problem, test the data, evaluate the prototype and preserve accountable human decisions.
Positioning
Clinical experience should shape the technology brief.
Football departments do not benefit from “AI” as a label. They benefit when a defined operational problem, suitable data and a safe decision process are matched to the simplest useful technology.
The work described here concerns responsible development and evaluation support. It is not an offer of automated diagnosis, autonomous treatment decisions or an approved medical device.
Possible work
From use case to an inspectable prototype.
02
Data readiness
Review provenance, missingness, labels, leakage risks, representativeness and permissions before model development.
03
Prototype and evaluation
Build or review transparent prototypes with temporal separation, relevant comparators and performance reported with uncertainty.
04
Governance and handover
Document limitations, human oversight, monitoring, data protection and the regulatory questions requiring specialist review.
Evidence and status
Training, software and research are not the same claim.
AI in Health Care
Harvard Medical School professional education listed in the public biography. Training supports informed development; it is not a product validation or regulatory approval.
Football Performance Dashboard
A limited offline workflow demo for squad availability and load metrics. It is software, not artificial intelligence and not a clinical record system.
Review the demo and its limitsModel-development prototypes
Research work may be presented only with dataset provenance, evaluation design, metrics and limitations. No experimental model on this site is represented as clinically validated.
Minimum project gates
What must be clear before deployment.
- Intended purpose, users and decisions.
- Lawful access to appropriately governed data.
- Baseline comparison and leakage-controlled evaluation.
- Failure modes, subgroup limitations and human escalation.
- Security, privacy, monitoring and change control.
- Independent legal, regulatory and clinical review where applicable.
Project enquiries
Start with the problem, not a model.
Describe the workflow, intended users and the decision you want to improve. Do not send health records, player identities, datasets or confidential club information through this form.
Development boundary: a prototype that diagnoses, predicts or recommends patient-specific clinical action may fall within medical-device and data-protection requirements. Classification and deployment require case-specific specialist review; this page does not claim regulatory approval.