Blaž Rolih
PhD Candidate
Škofja Loka, Občina Škofja Loka, Slovenia
Summary
🤩
RockstarBlaž Rolih is a PhD candidate in computer science at the University of Ljubljana with five years of practical experience in computer vision, remote sensing and anomaly detection. He contributes to production-ready open-source tooling as an ML engineer on the anomalib project, where he improved model performance, fixed OpenVINO inference issues, and added robust tiling and ensemble support for edge deployment. His work shows a strong focus on bridging research and engineering—tackling metric correctness, shape/type bugs, and replacing inefficient components to make state-of-the-art algorithms practical. Based in Škofja Loka, Slovenia, he combines academic rigor with hands-on optimization for real-world constraints, particularly on edge inference. Less obvious: his contributions reveal a talent for debugging cross-framework integration problems that often block model deployment.
5 years of coding experience
Github Skills (11)
computer-vision10
machine-learning10
pytorch10
anomaly-detection10
openvino10
model-optimization9
python9
tensorflow9
unsupervised-learning9
tensor9
testing8
Programming languages (4)
Github contributions (5)
open-edge-platform/anomalib
Mar 2023 - Mar 2025
An anomaly detection library comprising state-of-the-art algorithms and features such as experiment management, hyper-parameter optimization, and edge inference.
Role in this project:
ML Engineer
ML EngineerContributions:92 reviews, 19 PRs, 391 comments in 2 years 1 month
Contributions summary:Blaž primarily contributed to the refinement and optimization of the anomaly detection library. Their work focused on improving model performance and functionality, including fixing issues related to OpenVINO inferencing, correcting benchmark tests, and replacing components like `cdist` in the Patchcore model. Additionally, the user implemented and addressed shape and type issues within the PRO metric, as well as tackled tiling problems for models. They also added support for tiling for ensemble methods, and fixed transforms for several models.
hyper-parameter-optimizationstate-of-the-arthyperdetection-libraryexperiment-management
blaz-r/ML-For-Beginners
Aug 2021 - Mar 2022
12 weeks, 25 lessons, 50 quizzes, classic Machine Learning for all
Contributions:2 PRs, 46 pushes in 7 months
quizzespythonmachine-learningdata-science
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.