Jozef Mokry
Member Of Technical Staff at Cohere
Switzerland
Summary
🤩
Rockstar🎓
Top SchoolJozef Mokry is a Member of Technical Staff at Cohere with a decade of engineering experience building and optimizing machine learning systems and performance-critical infrastructure. He has a strong ML and NLP background from roles at PolyAI, Facebook, and internships at Google, and contributed to notable open-source projects like Nematus (neural machine translation) and KenLM (high-performance language modeling), where he implemented training improvements, layer normalization, beam search tweaks, and multi-threaded benchmarking. His work spans both model-level innovations and low-level performance engineering, reflecting comfort moving between research code and production-grade optimization. Cambridge- and Edinburgh-trained, he pairs rigorous academic foundations in computer science and data science with a pragmatic focus on shipping efficient, scalable ML systems. A less obvious strength is his history of refactoring research code into maintainable class-based designs, helping bridge prototypes to production.
10 years of coding experience
8 years of employment as a software developer
CDT in Data Science, CDT in Data Science at The University of EdinburghComputer Science BA, Computer Science BA at University of CambridgeGymnazium Ladislava SaruIB Diploma, IB Diploma at Gymnazium Jura Hronca
English, French, German, Slovak
Stackoverflow
13reputation
292reached
0answers
1question
Github Skills (19)
machine-translation10
c-language10
neural-machine-translation10
python10
multithreading10
benchmarking10
nmt10
tensorflow10
hashtable10
cprogramming-language10
algorithm9
algorithms9
normalizing8
boost8
normalization8
Programming languages (3)
Github contributions (5)
EdinburghNLP/nematus
Nov 2016 - Dec 2017
Open-Source Neural Machine Translation in Tensorflow
Role in this project:
ML Engineer
ML EngineerContributions:69 commits, 6 PRs, 11 pushes in 1 year 1 month
Contributions summary:Jozef's commits focused on enhancing the `nematus` project, an open-source neural machine translation system. Their contributions involved modifications to the training process, including saving and reloading optimizer parameters, and implementing layer normalization and other model architecture changes. Furthermore, the user refactored the code into classes, and worked on beam search and sampling.
machine-translationtranslationnmttensorflowsequence-to-sequence
kpu/kenlm
Jan 2016 - Feb 2016
Role in this project:Performance Engineer
Performance EngineerContributions:5 commits, 3 pushes in 18 days
Contributions summary:Jozef focused on performance benchmarking and optimization of probing hash tables within the KenLM project. Their commits introduce a multi-threaded prefetching benchmark to measure CPU time, and added an option for multiple tasks per thread, further enhancing the performance testing capabilities. These changes involved modifying and extending the existing benchmark code, primarily focusing on the `probing_hash_table_benchmark_main.cc` file to incorporate these new features.
nlplanguage-modelsmallerkenlmfaster
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.