Nikolay Bashlykov
Machine Learning Engineer, Meta AI
London, England, United Kingdom
Summary
🤩
Rockstar🎓
Top SchoolNikolay Bashlykov is a Machine Learning Engineer at Meta AI with a decade of experience building and scaling ML systems, currently focused on research and development of LLMs and distributed training across thousands of GPUs. He was a key contributor to Llama2 and Llama2-Chat pretraining and data-mix efforts, and has hands-on expertise stabilizing model-parallel training in MetaSeq for large-scale deployments. Previously he led data science teams and delivered measurable business impact—driving revenue uplifts, marketplace efficiency gains, and automated ML pipelines—while mentoring hundreds of students as an ML lecturer. Endorsed as UK Exceptional Global Talent, he combines deep quantitative roots from MSU and business training from Universität Mannheim with production-grade engineering, often tackling GPU memory and throughput bottlenecks that sit between research and reliable large-scale training.
9 years of coding experience
8 years of employment as a software developer
Master's degree, Business Administration and Management, Master's degree, Business Administration and Management at Universität MannheimLomonosov Moscow State University
Russian, English, French
Stackoverflow
1reputation
0reached
0answers
0questions
Github Skills (8)
pytorch10
machine-learning10
distributed-training10
parallelization10
python10
testing10
repr9
rep9
Programming languages (1)
Github contributions (5)
facebookresearch/metaseq
Jan 2023 - Jan 2023
Repo for external large-scale work
Role in this project:
ML Engineer
ML EngineerContributions:10 reviews, 1 commit, 18 PRs in 1 day
Contributions summary:Nikolay primarily contributed to testing and integrating model-parallel training configurations within the MetaSeq framework. Their work involved creating and modifying tests for model-parallel training, specifically for MP1 and MP2 configurations, ensuring correct training step counts and loss values. They also experimented with various configurations and parameters, and addressed issues related to CUDA memory, contributing to the stability and functionality of distributed training capabilities.
transformersdockerscalelarge-scalehuggingface
bashnick/transformer
Jan 2023 - Feb 2023
A codebase implementing a simple GPT-like model from scratch based on the Attention is All You Need paper.
Contributions:1 review, 19 commits, 1 PR in 1 month
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.