Since March 2017 I’ve held both the Software Architect and System Architect roles at Royal Philips, working across the company’s global medical-device portfolio — MR and CT scanners, defibrillators, patient monitors — in the strictly regulated, safety-critical world of medical devices. The work has spanned architecture, product ownership, risk management and a good deal of leading teams and deliverables. Lately my favourite thread has been System Models in Code: a small domain-specific language and compiler of my own design (SBDL) that expresses system specifications — and their FMEA — directly in the codebase, now in production at department scale and quietly underpinning an AI-augmented safety-analysis pipeline. These days I’m mostly architecting next-generation patient monitors, including cross-cutting functions like distributed, real-time analysis of patient vital signs across some ten engineering teams.

Sometime in the middle of 2015 I decided it’d be a good idea to try my hand in a larger-scale high-tech organisation, and in the Netherlands ASML is the logical choice. I spent almost two years there working on optical-calibration algorithms — a customer-critical, ~100k-line subsystem sitting inside one of the largest codebases in the world — and leading the introduction of Agile (Scrum, cross-functional teams) to the traditionally-structured metrology domain teams; a change that, I’m told, has stuck ever since.

In late 2010 I shifted areas and moved into medical image analysis. Based at DIAG in Radboud UMC, I worked in a dual capacity — both developing and researching the tools of the field, which centres on applying algorithms (increasingly machine-learning and neural networks) to medical image data like MRI, in order to help clinicians catch various disorders at an early stage. From late 2011 I focused on turning that research into quality-assured medical devices fit for commercial use — the engineering, certification and quality-management side — largely within the spin-offs DIAG Technologies and ScreenPoint Medical. Highlights include leading the software for ProCAD, a prostate-cancer diagnosis workstation, and the quality lead role for ScreenPoint’s mammography cancer-detection software; both neural-network-based, and both certified (CE / ISO 13485) for clinical use. In late 2012, in my spare time outside of work, I started writing the Simulacrum library and the Symmetry application. You can have a look at them here.

After my Ph.D. I joined the ‘Computer Systems Architecture’ (CSA) group at the University of Amsterdam (UvA), working on the compiler — a GCC 4 fork — for the parallel μTC language, a system-level language for the SVP (Self-adaptive Virtual Processor) model of concurrency, as part of the EU-funded AppleCore project. I also hand-coded a number of supercomputing benchmarks for evaluating the architecture. From 2009 I designed and built a novel high-speed, scalable parallel I/O system — together with a co-designed OS stack — for the massively parallel Microgrid many-core processor at the heart of AppleCore.

I earned my Ph.D. in the field of computer architecture and energy efficiency — thesis: ‘Energy Efficient Branch Prediction’ (Herts, England 2008) — demonstrating how to save significant processor energy by combining compilation and profiling techniques with only minimal hardware changes. Plenty of hands-on work at the hardware/software interface along the way: custom GCC tooling, a superscalar processor simulator with power modelling, and a kernel-level instruction tracer, all in C/C++ on Linux. If you’re interested in reading any of my previous research publications, you can still find many of them here.