MSc student in Computer Science at the Faculty of Electrical Engineering and Computing (FER), University of Zagreb. I work on computer vision, machine learning and GPU computing, and I am currently focused on accelerating mathematical morphology.

I like building the pieces most people import: Flash-style attention kernels, mathematical morphology libs,a PAM module for face login, etc.

Focus

Languages
Python · C++ · CUDA · Rust · Kotlin
ML & vision
PyTorch · NumPy · OpenCV
Systems
Linux · CMake · Meson · Docker · CI/CD

Projects

Sep 2026 – present

Python · Linux

bt-dualboot

Syncs Bluetooth pairing keys between Linux and Windows so paired devices survive a dual-boot switch.

Jul 2026 – present

Kotlin · IntelliJ Platform

IssueHub

JetBrains IDE plugin for working with GitHub and GitLab issues without leaving the editor.

Jun 2026 – present

CUDA · C++ · PyTorch

Morphottention

A PyTorch attention module built on mathematical morphology rather than the dot product. Flash-attention-style CUDA kernels for the forward and backward pass keep memory low by never materialising the full attention matrix; the module is autograd-differentiable and drops into a ViT unchanged. MIT-licensed, published on PyPI.

Jun 2026 – present

PyTorch · CUDA

serron

Mathematical morphology for PyTorch: differentiable erosion and dilation operators and learnable morphological layers, backed by hand-written CUDA kernels so they stay cheap to train with.

Mar 2026 – present

Research · PyTorch

Lesion Segmentation in Dermoscopic Images Using FCN with Mathematical Morphology Enhancements

Learnable mathematical morphology built into a fully convolutional network for skin-lesion segmentation: differentiable morphological layers, a boundary-aware loss, and morphological post-processing that repair the fragmented masks and lost thin structures plain FCNs produce. On ISIC2018-T1 it outperforms the original challenge submissions and later FCN methods, as well as several heavier U-Net and Transformer models, at a fraction of the compute. I led a five-person team and owned the morphological layers, the loss, preprocessing and evaluation.

Paper awaiting publication · code on acceptance

Mar 2026 – present

Python · C++ · PAM

SecureEye

Face authentication for Linux, wired directly into PAM: unlock the session and approve sudo with your face, with CLI tools to enrol and manage users. Python and C++, GPL-2.0; a ground-up take on the idea behind Howdy, built for current distributions.

Oct 2025 – Jan 2026

Django REST · React · PostgreSQL

BullBear

Full-stack paper-trading platform for stocks and ETFs: live market data, real-time portfolio tracking, return leaderboards and shareable “Mini-Funds”, built by a six-person team for FER’s Software Engineering course. As team lead I worked across the stack and owned the CI/CD pipeline, deployment, the database, and a cycle-based caching layer that batches market-data requests to stay under third-party rate limits.

Education

2026 – 2028

University of Zagreb, Faculty of Electrical Engineering and Computing (FER)

MSc, Computer Science (in progress).

2023 – 2026

University of Zagreb, Faculty of Electrical Engineering and Computing (FER)

BSc, Computer Science.

2019 – 2023

MIOC (XV. Gymnasium), Zagreb

High-school diploma, Mathematics & Computer Science.

Linux & Open Source Club · competitive-programming team (InfoKup, HONI, ACSL).

Languages

  • CroatianNative / bilingual
  • EnglishFull professional proficiency

Elsewhere