Blair Ferguson
AI-native systems for high-stakes decisions.
I build decision intelligence systems for clinical and regulated settings, and keep their decisions defensible when a clinician, regulator or board asks how they were reached.
What I do
Decision intelligence platforms
Modelling a decision explicitly, including its inputs, its flow, its owner and its escalation path, then building the platform that orchestrates it at scale and monitors the quality of what it produces.
Replatforming legacy estates
Most enterprises cannot bolt AI onto the systems they already run. I rebuild the foundations so that intelligence belongs to the architecture rather than sitting above it as a layer of automation.
Applied machine learning in high-stakes domains
Clinical and financial settings in which error is costly and an answer nobody can explain is worth nothing: computer vision on medical imaging, and risk and advice models in regulated finance.
Accelerated AI-native delivery
An AI-assisted development lifecycle that keeps its gates, reviews and traceability intact, so that delivering faster does not cost you the audit trail.
Three things I believe about putting AI into regulated work
Most AI failures are decision failures, not model failures
The model is usually sound. What tends to be missing is a clear definition of the decision it serves, an owner for that decision, and an agreed course of action for when the model is uncertain. Settle those and the accuracy problem usually recedes.
In regulated industries, explainability is the product
A recommendation nobody can justify carries risk rather than value, however good its metrics look. Treated as a requirement from the outset rather than a constraint bolted on late, explainability is what makes a system deployable at all.
Agents belong where the workflow is already understood
Autonomy applied to a process nobody has mapped produces untraceable errors quickly, not leverage. Understanding the decision first is unglamorous work, and it is most of the job.
Selected work
Preview Diagnostics
FounderDecision intelligence for medicine: machine learning for chronic disease detection from retinal scans, and hyperspectral imaging classification for intra-operative decision support.
- Retinal screening models: Machine learning for chronic disease detection from retinal scans, built so that a clinician can see why a case was flagged.
- Intra-operative decision support: Hyperspectral imaging classification that supports tissue decisions in theatre, where an unexplainable answer is worth nothing.
SCC Rugby Tracker
Built and operateA performance platform for a rugby academy, with multi-role access control, children's data handled under PDPA, and analysis the coaching staff genuinely use. Built and run single-handed.
How I got here
I hold a dual degree in Business Administration and Data & Business Analytics from IE University. Both halves matter equally: modelling earns its keep only when it changes what somebody decides, and a decision improves only when the modelling beneath it is sound.
Since then I have built and shipped production systems across advice-led financial services, clinical diagnostics and sports performance: decision platforms, computer vision on medical imaging, and the data foundations and access control that sit underneath both.
Current client engagements are not yet public, as the work is live and ongoing. I am glad to discuss specifics directly.
Get in touch
I am interested in problems where the decision matters more than the model, particularly in financial services and other regulated settings.