Volodymyr Halian
Linz, Austria · Focus: Machine Learning Engineering · volodiahalian@gmail.com
Education
Johannes Kepler University Linz
BSc Artificial Intelligence · Mar 2025 – Jun 2028 (expected)
Relevant coursework: Machine Learning, Pattern Classification, Unsupervised Techniques, Statistics for AI and algorithms.
Selected ML & AI projects
Sound Event Recognition — MLPC University Project
University MLPC project: prepared multi-label acoustic targets from annotations, used MFCC features with a PyTorch AudioMLP baseline, and structured recording-disjoint evaluation to reduce leakage.
Transformer Attention Lab — PyTorch (demo)
Implemented multi-head scaled dot-product attention, causal/padding masks and residual Transformer blocks in PyTorch; passed six deterministic checks of mask behavior and gradients on synthetic inputs.
Graph Neural Networks — GCN vs MLP (demo)
Implemented a two-layer GCN and feature-only MLP baseline; compared graph message passing with matched splits on a seeded synthetic, homophilic graph.
Research interests
Causal vs Full Attention — upcoming seminar and practical work at JKU Linz.
Transformer Attention Lab and GCN vs MLP are independently reproducible demos on synthetic data, not industry benchmarks.