Mike Chrzanowski
Research Scientist at NVIDIA
San Francisco, California, United States
Summary
🤩
Rockstar🎓
Top SchoolMike Chrzanowski is a research scientist with 14 years of experience building and deploying cutting-edge ML systems across industry leaders including NVIDIA, Google Brain, DeepMind, Baidu, and Adept. He specializes in applied deep learning and low-precision training techniques—most recently tackling fp8 pre- and post-training challenges at NVIDIA—bringing both research rigor and pragmatic engineering. His contributions to the high-profile NVIDIA/NeMo project include language-specific tokenization and preprocessing fixes for En-Ja machine translation, reflecting a knack for nuanced, language-aware data engineering. Mike holds an MS in Computer Science from Stanford and combines strong academic foundations with production-grade implementation experience. Colleagues value his ability to move ideas from prototype to robust system while debugging subtle pipeline and configuration issues. Based in San Francisco, he blends research curiosity with a proven track record of shipping improvements in large-scale generative and speech AI stacks.
13 years of coding experience
9 years of employment as a software developer
Bachelor of Arts - BA Computer Science Economics, Bachelor of Arts - BA Computer Science Economics at New York UniversityMaster of Science - MS Computer Science, Master of Science - MS Computer Science at Stanford University
Github Skills (8)
tokenize10
sentencepiece10
machine-translation10
tokenizer10
nlp10
python10
large-language-models9
encoder-decoder9
Programming languages (2)
Github contributions (5)
NVIDIA/NeMo
Feb 2021 - Jun 2021
A scalable generative AI framework built for researchers and developers working on Large Language Models, Multimodal, and Speech AI (Automatic Speech Recognition and Text-to-Speech)
Role in this project:
ML Engineer
ML EngineerContributions:16 reviews, 7 commits, 11 PRs in 4 months
Contributions summary:Mike primarily contributed to the machine translation components, focusing on Japanese and English language pairs. They implemented and refactored tokenization and detokenization processes using Moses, SentencePiece, and custom EnJa tokenizers, while also incorporating byte-level tokenization. Their work included debugging and fixing issues related to data preprocessing and model configuration, and integrating new features, as well as implementing configurations for language specific processors and tokenizers.
asrspeech-recognitionnatural-language-processingttsspeaker-diarization
mchrzanowski/libspatial
Jul 2014 - Jun 2016
Contributions:34 commits, 1 push in 2 years
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.