Marta Stępniewska-dziubińska
Software Engineer - AI at NVIDIA
Warsaw, Masovian Voivodeship, Poland
Summary
🤩
Rockstar🎓
Top SchoolMarta Stępniewska-Dziubińska is a Software Engineer specializing in AI with 14 years of experience, currently building scalable ML systems at NVIDIA from her base in Warsaw. She combines a strong academic background—a PhD in Biology and a Master’s in Bioinformatics—with hands-on MLOps and ML engineering, contributing to high-impact open-source projects like an AlphaFold 2 PyTorch reproduction and DeepChem. Her work spans optimizing training infrastructure, performance logging, and extending 3D deep learning layers for scientific applications, reflecting a rare blend of computational biology insight and production-focused engineering. Known for pragmatic problem-solving, she’s improved distributed training behavior and low-level model components that directly boost throughput and maintainability.
14 years of coding experience
3 years of employment as a software developer
Master's degree, Bioinformatics and Systems Biology, Master's degree, Bioinformatics and Systems Biology at Uniwersytet WarszawskiDoctor of Philosophy - PhD, Biology, Doctor of Philosophy - PhD, Biology at Institute of Biochemistry and Biophysics PAS
Github Skills (12)
pytorch10
machine-learning10
deepspeed10
deep-learning10
tensorflow10
performance-optimization10
python10
unit-test10
drug-discovery9
quantum-chemistry8
devops8
science8
Programming languages (4)
Github contributions (5)
deepchem/deepchem
Sep 2017 - Nov 2017
Democratizing Deep-Learning for Drug Discovery, Quantum Chemistry, Materials Science and Biology
Role in this project:
ML Engineer & Data Scientist
ML Engineer & Data ScientistContributions:30 commits, 4 PRs, 15 comments in 1 month
Contributions summary:Marta primarily contributed to the implementation and testing of 3D convolutional layers and max pooling layers within the `deepchem` repository, a project focused on democratizing deep learning for scientific applications. They added the `Conv3D` and `MaxPool3D` layers, including their associated tests, demonstrating a focus on extending the library's capabilities for 3D data processing. Furthermore, the user was also responsible for fixing file format in the rdkit_grid_featurizer. These changes suggest a focus on enhancing the deep learning models within the domain of drug discovery, quantum chemistry, and materials science.
chemistrysciencedrug-discoverydeep-learningbiology
aqlaboratory/openfold
Oct 2021 - Nov 2021
Trainable, memory-efficient, and GPU-friendly PyTorch reproduction of AlphaFold 2
Role in this project:MLOps Engineer
MLOps EngineerContributions:18 commits, 1 branch in 12 days
Contributions summary:Marta focused on improving the performance and maintainability of the training process for the AlphaFold 2 reproduction. Their contributions include adding a performance logging callback for measuring throughput and latency, optimizing the loss function, and correcting a bug in argument parsing. They also made adjustments to the training infrastructure by modifying the use of DDP based on the number of GPUs and nodes. They also updated the logging configuration using dllogger.
pytorchmemorygpualphafoldreproduction
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.