Nikolay Bogoychev
Research Scientist at Meta
City of Edinburgh, Scotland, United Kingdom
Summary
🤩
Rockstar🎓
Top SchoolNikolay Bogoychev is a research scientist and performance-focused engineer with 13 years of experience building and optimizing neural machine translation systems and HPC workloads. He has driven production-grade int8 CPU/GPU inference and tensor-core optimizations as a major contributor to Marian NMT and built on-device, privacy-first translation solutions as CTO of Efficient Translation. At Edinburgh and Meta he combined deep NLP expertise (transformers, transfer learning, embeddings) with low-level GPGPU/CPU GEMM tuning and benchmarking to make translation faster and more deployable. A competitive programmer at heart, he enjoys algorithmic problem solving and brings that creativity to systems-level optimization and tooling. He is an active open-source contributor with patches across Marian and Moses and a long-standing commitment to reproducible, high-performance ML engineering. Unusually for an ML researcher, he also studies logographic writing systems, feeding linguistic curiosity back into his modeling work.
13 years of coding experience
6 years of employment as a software developer
Doctor of Philosophy (PhD), Informatics, Doctor of Philosophy (PhD), Informatics at The University of EdinburghHigh school, Spanish Language, 5.91, High school, Spanish Language, 5.91 at 164 Spanish Language High School
Bulgarian, English, Spanish, German, Chinese
Stackoverflow
1,462reputation
148kreached
46answers
28questions
Github Skills (17)
machine-translation10
c-language10
neural-machine-translation10
python10
gpu10
cuda10
cprogramming-language10
testing9
test-framework9
optimization9
deep-learning8
git8
profiling7
compression6
constexpr6
Programming languages (14)
Github contributions (5)
moses-smt/mosesdecoder
Jun 2014 - May 2016
Moses, the machine translation system
Role in this project:
Back-end Developer
Back-end DeveloperContributions:184 commits, 33 pushes, 2 branches in 1 year 11 months
Contributions summary:Nikolay implemented a speedtesting framework for the Moses machine translation system, including the framework itself and related documentation. The user's work primarily involved creating a Python-based testing suite, configuring repository initialization, and implementing log writing functionalities. Furthermore, the user contributed to the analysis of test results, generating HTML reports, and integrating the code with Google Profiler.
i18nmachine-translationtranslationlocalizationmoses
marian-nmt/marian
Jul 2017 - Dec 2021
Fast Neural Machine Translation in C++
Role in this project:Back-end Developer
Back-end DeveloperContributions:49 commits, 4 comments, 2 issues in 4 years 5 months
Contributions summary:Nikolay primarily contributed to the optimization and maintenance of the Marian NMT project. They addressed issues related to memory management within the CUDA environment, as seen in the `nth_element.cu`, `sparse.h`, and `training/dropper.h` file changes. Furthermore, the user worked on improving the optimizer's handling of larger batches, specifically within the `optimizers/optimizers.h` and `optimizers/optimizers.cu` files. They added command line options for flexible learning rates, and updated the graph group to allow for dynamic learning rate with tau.
machine-translationtranslationneural-machine-translationcudagpu
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.