Computer Vision · Mathematics · Surgical AI

Shahnawaz “Shah” Ahmed

Co-Founder & CTO, Scalpel AI · PhD Computer Vision · London

I build the intelligence layer that makes every step in surgery observable and accountable.

Geometry and machine perception, aimed at a single goal: teaching the systems around surgery to see, so that nothing in a patient’s care is left to luck or memory.

Shahnawaz Ahmed

Fields of Focus

Surgical IntelligenceAI InfrastructureGovernance & ReliabilityPerception Under PressureLearning DynamicsGeometric Modelling

01 · The Problem

Surgery is medicine at its most unforgiving.

Yet its safety still rests on fragmented systems, manual counts, and the corner of a practised eye.

02 · The Work

The Work

Geometry & Perception

Seeing structure where machines fail

My work sits at the intersection of geometry, perception, and systems engineering. During my PhD at Queen Mary University of London, I focused on one of computer vision’s hardest challenges: identifying and matching patterns in low-textured environments.

When there are no obvious edges or clean features, machines struggle: appearance stops being evidence. My thesis answered with mathematics: structure recovered precisely enough to be acted on, even where the surface gives nothing away.

Connected Surgical Intelligence

Making surgery measurable

At Scalpel AI, I lead the engineering and AI teams turning surgery, and the supply chain behind it, into a measurable, verifiable system.

By combining computer vision, machine learning, and multimodal sensing, we analyse workflows in real time, detect abnormalities, verify protocol compliance, and track surgical instruments automatically. The goal has never changed: no patient harmed by an error a machine could have caught.

Verifying surgical instruments is not so different.

An instrument tray is that problem, sharpened. Most instruments are polished steel, so to a camera each is a mirror wearing the room’s reflection; many are near-twins, identical except for a millimetre of length or width; and the open web has barely photographed this world. So identity is read from geometry: shape, proportion and structure, checked against imagery, models and ground truth we built ourselves.

03 · The Supply Chain

Hours before the first incision

Behind every operation is a logistics operation. It begins hours before the first incision: in a warehouse, where instruments are counted into trays; on the road, where vans thread between facilities; and at every hand-off, where a human being checks the contents by eye. The system is expected to get every tray there complete, sterile and exactly right, thousands of times a day. Mostly, it does.

“Mostly” is the problem. A wrong tray does not announce itself at the warehouse or on the motorway. It waits, and surfaces at the worst possible moment: in theatre, with the patient asleep on the table. A missing instrument stalls the case; a wrong one introduces a risk nobody consented to. And the failure is almost never carelessness. It is architecture: the chain cannot see itself.

What sight the chain does have is borrowed from people: the veteran who assembles a forty-instrument tray from memory, the coordinator who knows three hospitals call the same clamp by three different names. That knowledge lives in heads, and heads retire, resign, go on holiday. Scalpel AI gives the chain its own eyes: machine vision that verifies every tray, instrument by instrument, at the warehouse, in the field and at the hospital, so an error is caught where it is made, and what once lived in memory is written into the system.

First-inspection errors · 10,853 joint reconstruction trays · three months

fell from 4.3% to 0.2%

Measured with a US surgical logistics provider. The full case study(opens PDF in new tab) is public at scalpel.ai(opens in new tab).

Sight changes the map. A tray verified in the field moves straight to the next hospital instead of going home first: the same trays serve more cases, the same instruments more patients, and the freight bill stops paying for empty miles.

This is Connected Surgical Intelligence.

Every instrument is tracked.

Every step is validated.

Every site sees the same truth.

05 · Philosophy

Building AI for healthcare is not about novelty.
It is about trust.

A patient under anaesthesia cannot check our work. That is what trust means here: systems precise enough to be proven, robust enough to be boring, and accountable enough that someone signs their name to every decision.

If your system has to be right at 3 AM, we should talk.