Forward-deployed AI engineering
Probability Mass is a boutique AI engineering practice. We embed in your team and ship the AI work that matters into your real systems - model-agnostic, hands-on, and accountable for the outcome.
What we do
We take the AI feature or process that matters most and put it into production inside your existing systems. No rip-and-replace, no proof of concept that dies in a slide deck.
We own the AI roadmap and sit in the exec conversation - technical strategy, build-vs-buy, hiring - and carry the accountability that goes with it.
Clinical AI, EU AI Act, safety guardrails, clinical-grade data governance across jurisdictions. Where being wrong is expensive, we de-risk the integration.
We optimise for you, not for our majority shareholder.
Proof
500,000+
A 10-year risk score derived from over 500,000 individuals, requiring no blood test and no blood-pressure measurement. Built with Bayer Consumer Health, released in the US in 2023 to reach more than 100 million at-risk Americans, and extended to Saudi Arabia in 2024. Methods published in European Heart Journal - Digital Health and validated against the US All of Us dataset.
$6M+
Co-founded a London AI startup predicting mental health from wearable data, raised over $6M, and led it to acquisition in 2020 - then directed the AI organisation inside the acquirer across digital biomarkers, cardiovascular, oncology and mental health.
24/7
A GPU and CPU model-serving cluster, MLOps and deployment pipelines, and PACS integration running computer-vision models in production for veterinary radiology. Co-inventor on the resulting patents.
NHS
Generative AI shipped into production for clinicians and operations teams across a dual UK and US mental health service - clinical note translation, NHS-ready reporting, automated report audit and conversational clinical support.
20+
Published across European Heart Journal - Digital Health, Scientific Reports, JMIR, Sensors and the Journal of Artificial Intelligence Research, with granted patents in biometric signal processing.
Selected past clients and institutions
Current engagements: a UK digital health platform, a VC group, and a veterinary imaging AI company.
Who you work with
Two decades founding and leading technical teams from zero to acquisition, deploying production AI for clinical use, and doing academic research at the University of Oxford.
Two doctorates - Computer Science at Pisa and a DPhil in Biomedical Engineering at Oxford, where he is a Research Associate. Has lectured at Oxford, Surrey, Pisa and IMT Lucca.