What I believe

Most AR HUD commentary treats the display as the product. It isn’t. An AR head-up display is a perception problem disguised as a display problem.

Projecting graphics onto a windshield is well-understood optics. Projecting the right graphics, in the right place, adjusted for the driver’s eye position, compensating for windshield curvature and refraction, under changing light and weather, at latencies measured in milliseconds, in a system that has to meet automotive safety certification – that’s a different engineering challenge entirely. It’s a real-time perception-display loop, and the perception side is where the unsolved problems live.

I’ve spent my career at this intersection: computer vision, tracking, sensor fusion, 3D graphics, real-time systems, and optical calibration. Most people building AR HUDs come from one of those disciplines. Few have shipped production systems in all of them. That experience shapes how I think about the problem.

The edge is the product.

In automotive or defence, you don’t get to phone home. Your system has to perceive, decide, and render within a latency budget that makes cloud processing irrelevant. I started my career building tracking systems on hardware with a fraction of the compute we have today – constrained environments were the norm, not the exception. That mindset matters more in AR than most people appreciate.

Demos show intent, not capability.

I’ve shipped dozens of systems that started as impressive demos and became painful the moment they met weather, procurement, and people. The transition from prototype to product is where most deep-tech companies fail – and it’s where I’ve spent most of my working life. If a demo can’t explain what breaks at scale, it isn’t a demo. It’s a wish.

Measurement beats intuition.
This applies to everything from eye-tracking accuracy to engineering team health. If you can’t define what “good” looks like in numbers, you can’t improve it, and you certainly can’t ship it. I got this from working alongside professional athletes, where marginal gains decide careers and refusing to measure is a competitive disadvantage.

25 years of shipping perception systems

2024 – present

Co-founder

Distance Technologies

  • – Co-shaped the software architecture and helped build the initial R&D organisation across graphics, computer vision, embedded, and tooling
  • – Shipped 10+ customer and internal projects on time and to spec, including hands-on development of face-tracking for early 3D AR HUD prototypes
  • – Co-inventor on 13+ patents; supported technical due diligence for ~€13M raise (GV, Maki VC, FOV Ventures)

2018 – 2023

Principal CV Researcher → Senior Director, Device Software

Varjo

  • – Led hands-on C++ development of eye tracking for VR-1, Varjo’s first commercial headset
  • – Helped scale device-side R&D from ~10 to ~70 developers across several headset generations (VR-1, XR-1, VR-2, XR-3, VR-3, Aero, XR-4)
  • – Built teams delivering eye tracking, inside-out tracking (SLAM + IMU fusion), AR anchors, calibration & QC tooling
  • – Named inventor on 4 granted US patents in gaze tracking and XR display systems

2002 – 2013

Technical Director & Co-founder

Venatrack / Real Time Tracking – acquired by Sportradar AG, 2011

  • – Built pioneering multi-object optical tracking: speedway (2003), greyhounds (2004), Premier League football (2007)
  • – Delivered world’s first international broadcast of 3D pixel-based player-motion graphics (2009 FA Cup Final)
  • – Player statistics and analytics broadcast to 205 countries (Premier League Productions, 2012)
  • – Twice winner of DTI SMART Awards for innovation in machine-vision perception

1997 – 2017

Independent work & earlier ventures

Twilight 3D, Incagold, Independent consulting

  • – Pit-lane exit monitoring system for Formula 1 (Formula One Management)
  • – Co-authored 3DGE game engine – 20+ published titles, demos for Matrox G200 and Bitboys GPUs
  • – Interactive installations for the Finnish Institute in Paris, Kiasma, and Forum Marinum
  • – CTO for a PC/console game publisher (Zurich) – technology strategy and IPO preparation

Principles I return to

Build for the constraint, not the spec sheet.

Latency budgets, thermal envelopes, cost targets, and certification requirements define the real engineering problem more than the feature list does. The constraint is what makes the problem interesting – and what makes the solution defensible.

Prototype fast, productise slowly.

Speed matters in the early stages: get signal, prove feasibility, kill bad ideas cheaply. But the transition to production is a different discipline – testing, reliability, manufacturing, environmental variation. Most of the effort lives here, and most of the failures are from teams that underestimate it.

Score goals as a team.

One of the best things about technical work is that almost anyone can score goals – a junior developer who solves a blocker, a QA engineer who catches a critical bug early, an intern who spots a pattern nobody else saw. The job of leadership is to create the conditions where that happens regularly, not to personally do all the scoring.

Live ahead of the org.

Whoever runs engineering should be thinking six months ahead, anticipating hiring, architecture decisions, and problems nobody has complained about yet. Technical leads should live two months ahead, translating strategy into concrete plans. Engineers should live a sprint or two ahead, focused on execution. This only works if leadership has genuine technical depth – you have to see the curve balls coming.

Intellectual Property

17+
Patents across gaze tracking, XR display systems, and AR HUD perception, tracking, and display – spanning work at Distance Technologies (13+ granted/published patents, 10+ pending) and Varjo (4 granted US patents, plus international filings).