Nenad Markuš

This website contains basic info about myself, my research interests and links to some of my projects.

Work experience

2025–present	AI Technology R&D and Technical Leadership — overseeing software development, infrastructure, hardware, and technology strategy across the ONCE SPORT ecosystem
2023–2025	Leading Researcher on an EU Defence Fund project at the University of Zagreb, Faculty of Electrical Engineering and Computing
2019–2023	Freelance work for various EU and US companies, founding and work on Autocam^*
2018–2019	Microblink Ltd.
2011–2018	Teaching and research assistant at the University of Zagreb, Faculty of Electrical Engineering and Computing

A (partial) list of companies I worked with:

References can be provided on request.

^* Autocam is an automated recording solution that leverages AI technology to create TV-like footage from panoramic sports videos recorded with a GoPro or other wide-angle cameras.

I co-founded the project in 2019 in partnership with ONCE SPORT d.o.o., initially taking responsibility for its development and technology, while ONCE SPORT handled marketing, sales, customer support, and other business operations.

Since then, my involvement has expanded to encompass the broader development of the Once Sport technology stack, alongside my role as a shareholder and director of the company.

Rather than establishing a separate company to hold the Autocam intellectual property, we integrated the project into ONCE SPORT, creating a long-term structure that combines our respective technology and business capabilities.

Autocam has been an ongoing project since 2019 and continues to gain traction among users.

Research

During my university years, I did research, R&D and consulting in the following areas:

face analysis (face tracking, eye pupil localization, gender/age estimation, etc.);
real-time object detection;
non-photorealistic rendering (image mosaics, ASCII art);
learning local descriptors for keypoint matching and image recognition;
neural networks and deep learning in general.

At Microblink, I mainly worked on problems in vision-based document processing/recognition/retrieval on mobile platforms.

However, I am interested in a variety of topics in machine learning, computer vision and computer graphics, and plan to do research in these areas as well.

Selected publications

N. Roso, S. Srebot, N. Markuš, M. Sužnjević
When NPCs take their time: Token latency effects in LLM-driven game conversations, Entertainment Computing, 2026
[paper]

N. Roso, S. Vlahović, N. Markuš, M. Sužnjević
Exploring Conversations with AI NPCs: The Impact of Token Latency on QoE and Player Experience in a Text-Based Game, QoMEX, 2024
[paper]

N. Markuš, M. Sužnjević
Theoretical and Empirical Analysis of a Fast Algorithm for Extracting Polygons from Signed Distance Bounds, Algorithms, 2024
[paper][project]

N. Markuš, I. S. Pandžić, J. Ahlberg
Learning Local Descriptors by Optimizing the Keypoint-Correspondence Criterion: Applications to Face Matching, Learning from Unlabeled Videos and 3D-Shape Retrieval, IEEE Transactions on Image Processing, 2019
[paper][project]

N. Markuš, I. Gogić, I. S. Pandžić, J. Ahlberg
Memory-Efficient Global Refinement of Decision-Tree Ensembles and its Application to Face Alignment, BMVC, 2018
[paper]

N. Markuš, M. Fratarcangeli, I. S. Pandžić, J. Ahlberg
Fast Rendering of Image Mosaics and ASCII Art, Computer Graphics Forum, 2015
[paper][project]

N. Markuš, M. Frljak, I. S. Pandžić, J. Ahlberg, R. Forchheimer
Eye pupil localization with an ensemble of randomized trees, Pattern Recognition, 2014
[paper][demo]

N. Markuš, M. Frljak, I. S. Pandžić, J. Ahlberg, R. Forchheimer
Object Detection with Pixel Intensity Comparisons Organized in Decision Trees, technical report, 2013
[paper][project]

Research notes/posts

Read these here.

Projects and code

Please visit my github page.

`pico`

My most popular contribution to the open-source community is pico, a minimalistic face-detection engine suitable for deployment on mobile devices and embedded hardware. The core algorithm behind pico was developed during my PhD studies. Nowadays it is deployed on millions of devices, in both commercial and non-commercial projects.

There's also an implementation in JavaScript: picojs.

The open-source community has made several reimplementations of pico. The most popular one being probably pigo.

`LambdaCAD`

A tiny, in-browser CAD tool which enables you to specify 3D models by writing JavaScript code.

An in-browser demo is available here. The code can be found on GitHub: https://github.com/nenadmarkus/gridhopping.