Waqar Shah

Medical Doctor · Clinical AI Architecture · Safety · Governance

I turn complex clinical knowledge into safer, governable AI systems.

Applying rigorous standards to support AI implementation in healthcare. Helping healthcare organisations understand and demystify how they make decisions, to allow effective and safe AI implementation. Providing support and learning to clinicians working with AI.

CLAIMEVIDENCE

Page 3 is open. The chart is waiting.

  1. BeforePatient-facing triageGo
  2. DuringAmbient scribe evidenceGo
  3. AfterSummarising agentOpen

See how I think.

03 / After the consultation

The summarising agent

A patient chart, nine days long. An agent offers to read it for you.

Synthetic patient and chart, written to illustrate failure modes. Not a real record.

Ms A's chart: 68 documents across nine days.

01 /What I do

Four connected practices.

Understand your decisions. Govern your AI. Own your Alpha.

Clinical and knowledge architecture

Map how decisions are made and turn clinical reasoning into structures that teams and AI systems can use.

Clinical-AI safety and governance

Define intended use, hazards, controls, provenance, uncertainty, oversight and evaluation.

Digital clinical safety

Safety-management planning, hazard logs, incident processes and deployment gates for clinical digital systems — built on NHS digital clinical-safety training.

Product strategy and delivery

Move from clinical problem through discovery, architecture, implementation, evidence and adoption.

02 /Research

Patented thinking, honestly framed.

Co-inventor of a published UK patent application for a medical-imaging method: reconstructing three-dimensional vascular anatomy from two-dimensional ultrasound. My contribution spanned the initial concept, system and research architecture, methodology, model deployment and synthetic-data production.

A clean-room technical demonstration built entirely on synthetic and public data is in development for the Laboratory section of this site.

UK application
GB2416877.5
UK publication
GB2702377A

Clinical research · Pilot

The VividGen pilot

A University of Southampton and VividGen Ltd pilot, run entirely on synthetic data. Parametric vessels were cut into simulated ultrasound frames, a segmentation network was trained on those frames, and each held-out vessel was rebuilt in 3D and scored against the vessel it came from — once from clean frames and once from speckled ones.

Reading the pilot repository…

Open the pilot: results and visualiser
A vessel from the pilot's generator, drawn to scale. The teal square is the probe's 20 mm field of view.

03 /Availability

Selected opportunities.

Available for selected employment, advisory and collaboration opportunities in clinical-AI strategy, product architecture, synthetic evaluation, safety and governance.