skip to content
John Sietsma

About

I started out in artificial intelligence, which meant something different then. I helped commercialise a neural-network search engine that categorised documents and answered questions posed in plain English, for clients like Telstra and Intel. I followed it with a Master’s in intelligent systems, specialising in agent systems, and wrote a thesis on modelling emotions in non-player characters.

Then I moved into real-time 3D and spent a long stretch making games.

I led a technology team at Transmission Games, where we rebuilt London from OpenStreetMap data so players could fly across a hundred square kilometres of it. I led the programming team at Big Ant. I co-directed League of Geeks and architected Armello, a turn-based multiplayer game, largely single-handedly. I was the sole programmer on Oscura Lost Light.

Games taught me the thing I’ve used ever since: how to make real-time 3D work under hard constraints, and how to make a place feel like somewhere you could actually be.

Then I got interested in the real world getting into the frame. At Dekko, a San Francisco startup, I built augmented reality on iPad where virtual objects interacted with real surfaces, with no markers and no depth sensor, just the camera. At Many Monkeys I wired networks of Kinects together to reconstruct physical spaces as point clouds for installations. At the University of Newcastle I built VR learning experiences: a crime scene for criminology students, a radiation lab for physics students.

Somewhere in there I became Unity’s Technical Evangelist for Australia and New Zealand. We doubled the business year on year. I wrote core shaders that still ship in AR Foundation, helped put on a 500-person conference three years running, and taught shaders and rendering to everyone from primary schoolers to conference halls. I’ve taught game programming at university level and started an after-school games club at a primary school. Teaching is the part I’ve never wanted to stop doing.

More recently I’ve worked closer to the data. I led a team of six on rail infrastructure, processing point clouds at a scale I hadn’t worked at before. Then more machine learning: tracking players from broadcast footage, measuring athletic performance from a phone, and an acoustics platform that lets engineers annotate, train and publish their own models instead of waiting on someone like me. That last one became a paper at Inter-Noise 2026. It closed a loop I’d left open a long time ago.

What I keep coming back to is work where the thing being captured matters to someone.

So I’m building CultCap, a pipeline that turns handheld phone footage of heritage objects into Gaussian splats and archival-quality meshes. A museum shouldn’t need a $50,000 scanner to put an object online. Most collections hold millions of items and give each one a single flat photograph.

I’m in Newcastle, Australia. I write about what I learn, I publish the methodology, and I’m interested in work where the 3D is pointed at something worth keeping.

Publications

Clayton Sparke and John Sietsma. An Update on Evaluating Sound Pressure Levels of Noise Sources in Busy Noise Environments with Deep Learning. Proceedings of INTER-NOISE 2026, Adelaide, August 2026.

Patent applications

Elsewhere