I design and build interfaces to intelligent systems

For fifteen years that meant making machine perception legible to people — hand tracking, gaze, speech, ML-based sensing — in cars, headsets, wall-sized displays and public installations. Now it mostly means agents, LLMs and generative AI, which is a similar problem of non-deterministic input and output with lots of uncertainty. A tracker that misreads a gesture and a model that misreads an intent are a similar design problem — how sure is it, how does the person see that, and how do they take it back. Agents and LLMs just moved the guessing from perception to reasoning.

I think about the product, design the interaction and then build the thing itself (previously by hand, now with AI of course). With systems that behave probabilistically this lets me live with it for a while, know its ins and outs and improve it over time.

A few things I’ve made:

• Pixel Plaza — a permanent LED installation in the UK with several AI and generative experiences, plus a reusable generative AI workflow and UX for quickly spinning up new experiences (in days rather than weeks). 30k+ public interactions.
• A conversational AI POC in mixed reality for a top consumer brand (under NDA). I led UX for a voice-driven agent with spatial UI, deciding the agent behavior, the conversation flows and the right manual interactions to complement the voice input. Happy to walk through it privately.
• MBUX Interior Assistant — I led UX design and prototyping at Mercedes-Benz R&D for a set of in-cabin interactions driven by a depth camera and custom ML hand tracking. It went into production in the S-Class.
• Gesture, gaze and speech research at Intel RealSense — the work became the design guidelines external developers built their RealSense apps against.
• A multi-user, multi-modal interaction framework from my PhD, which ended up running at over 100 research and industry sites worldwide.
• AR LaserTag — 50k organic downloads, 7M views, zero marketing spend. The least serious thing on this list and possibly my favourite.

Four patents. Red Dot Best of the Best. Best of CES. A dozen peer-reviewed papers.

Right now

I’m freelancing on AI-driven experiences alongside the Hot Dark Matter work, and building my own AI tools. One example is Growthbot which is a personal growth agent that learns about me and helps me overcome my own limiting beliefs, fears and other mental blocks. It’s built on OpenClaw with local models — I’ve been using it and improving it daily since May 2026, which is the only true test of an interface.

During my PhD

in HCI at the Electronic Visualization Laboratory (University of Illinois at Chicago), I studied how ultra-high-resolution wall displays work as a shared medium for collocated group work, and built the multi-user, multi-modal interaction techniques that made it possible. The underlying question — how do several parties act on one shared surface at once without stepping on each other, and how does anyone tell who did what — turns out to have aged well.

I’ve worked at

Mercedes-Benz R&D, Intel RealSense and NASA Ames, all in Silicon Valley; built AR/VR experiences for Sony, Coca-Cola and Epic Games at Happy Finish in London; and since 2022 I’ve been co-founder and CTO at Hot Dark Matter, building geospatial AR, large-scale LED installations, generative-AI pipelines and conversational AI prototypes in mixed reality.

Away from screens

I’m in the mountains with a bike, skis or a camera. I spent years building, racing and hacking cars. Lately, lots of my free time is spent with my son, helping him figure out this whole life thing and learning how to help him stay himself.

Born in Zagreb, Croatia. Lived in Pennsylvania, Chicago, California, Barcelona and London. Currently in Lisbon, working remotely and open to relocating for the right thing.