Eldar Kurtić
Principal Research Scientist at Red Hat
Vienna, Austria
Summary
🤩
Rockstar🎓
Top SchoolEldar Kurtić is a Principal Research Scientist based in Vienna with 8 years of experience specializing in neural network compression and efficient inference for CPUs and GPUs. He has driven production-focused research and engineering at organizations including Red Hat, IST Austria, and Neural Magic, shrinking large models to run faster without sacrificing accuracy. His background spans computer vision for autonomous driving to LLM compression, combining hands-on C++, Python/PyTorch deployment work and algorithmic research. An active contributor to the Hugging Face Transformers codebase, he has made compatibility and activation-function improvements that reflect a pragmatic attention to detail in widely used ML tooling. Eldar holds a MicroMasters in Statistics and Data Science from MITx and an M.Eng. in Automatic Control and Electronics, pairing strong theoretical training with practical systems experience. Colleagues rely on him to bridge research prototypes and production constraints, especially where compute efficiency is paramount.
8 years of coding experience
4 years of employment as a software developer
MicroMasters Program in Statistics and Data Science, 98/100, MicroMasters Program in Statistics and Data Science, 98/100 at MITxMaster of Engineering (M.Eng.), Automatic Control and Electronics, Master of Engineering (M.Eng.), Automatic Control and Electronics at University of Sarajevo
Github Skills (9)
transformer10
machine-learning10
pytorch10
nlp10
deep-learning10
python10
natural-language-processing10
model-optimization9
documentation8
Programming languages (6)
Github contributions (5)
huggingface/transformers
May 2021 - Apr 2022
🤗 Transformers: State-of-the-art Machine Learning for Pytorch, TensorFlow, and JAX.
Role in this project:
ML Engineer
ML EngineerContributions:4 reviews, 5 commits, 5 PRs in 11 months
Contributions summary:Eldar contributed to the Hugging Face Transformers library by fixing a docstring typo, replacing `BertLayerNorm` with `LayerNorm` in a supporting example, and implementing activations as PyTorch modules. These changes demonstrate a focus on refining the code and improving compatibility within the framework. Furthermore, the user made updates to the MNLI example by preventing overwriting metrics. Finally, the user updated activation functions by using ACT2FN.
pythonbertspeech-recognitionstate-of-the-artflax
eldarkurtic/sparseml
Apr 2021 - Feb 2023
Libraries for applying sparsification recipes to neural networks with a few lines of code, enabling faster and smaller models
Contributions:126 commits, 146 pushes, 55 branches in 1 year 10 months
recipessmallerlinesdeep-learningmachinelearning
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.