Six moves from a fuzzy problem to a shipped decision.

Diagnose

Name the real problem before sketching a solution.

Research

Study the system, the data and the constraints already in the room.

Talk to humans

Sit with the people who will use it and the ones who will pay for it.

Light-bulb moment

The reframe that turns a constraint into the core idea.

Data & metrics

Make the signal legible so a decision can be defended.

Solution & adoption

Ship the thing, then make it stick in real workflows.

0of 6 stages visited

Selected work

This work is under NDA, so client screens stay private. Each case study below narrates the design process through original visuals drawn from scratch. Request for Password.

A correction surface for an 85% accurate pipeline NDA

An AI editor that turns uncertain model output into structured training signal.

Human in the loopTrust calibrationInternal ML enterprise

Scandit2023 – 24 · Sole IC

Making ML output actionable NDA

Turning raw model output into something an enterprise buyer could evaluate and an operator could act on.

Trust calibrationIAData Annotation & Editing

Scandit2023 – 25 · Sole IC

Activation failure was a growth constraint NDA

Making invisible system states visible, checkable and recoverable. Time to first scan cut from five minutes to two.

OnboardingState modellingSDK

Scandit2023 – 24 · Lead IC

Designing AI reliability for retail execution NDA

AR, OCR and computer vision for real-time price intelligence, rolled out with Staples Canada.

ARComputer vision

Scandit2022 – 24 · Lead IC

Public Work

Agentic Business Investigator