Yi Dong
Principal Research Scientist at NVIDIA
Greater Boston United States
Summary
🤩
Rockstar🎓
Top SchoolYi Dong is a Principal Research Scientist at NVIDIA with seven years of focused experience translating deep scientific training into production-grade AI systems. He holds a Ph.D. in Computational Neuroscience from Johns Hopkins and applies biologically inspired models to advance model alignment and reasoning, leading projects such as SteerLM, SteerLM2, and llama-3.1-nemotron-70b-instruct. At NVIDIA he has moved between research and applied roles, contributing to NeMo by integrating Megatron-based BERT models and ensuring scalable training and checkpoint compatibility for large-model workflows. His background spans physics, quantitative finance, and software engineering, enabling a rare blend of theoretical rigor and pragmatic engineering. Based in Greater Boston, he combines academic impact—documented on Google Scholar—with hands-on open-source contributions that bridge research prototypes and developer-ready frameworks.
7 years of coding experience
15 years of employment as a software developer
Doctor of Philosophy (Ph.D.) Computational Neuroscience, Doctor of Philosophy (Ph.D.) Computational Neuroscience at The Johns Hopkins University School of MedicineJohns Hopkins UniversityB.S. Physics, B.S. Physics at Nanjing University
English, Chinese
Github Skills (17)
parallelization10
machine-learning10
large-language-models10
ml10
deep-learning10
trainings10
bert10
neural-network10
modeling10
pytorch9
optimization8
generative-ai8
deep-q-learning8
asr5
tensorflow5
Programming languages (6)
Github contributions (5)
NVIDIA/NeMo
Dec 2021 - Jan 2023
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:411 reviews, 421 commits, 106 PRs in 1 year 1 month
Contributions summary:Yi's commits primarily focus on adding and modifying code related to integrating Megatron-based BERT models within the NeMo framework. Their work includes supporting Megatron-NeMo Bert models, fixing dataset issues, and verifying the training and parallelization of these models. They also contributed to converting Megatron LM checkpoints into NeMo format and updating tutorials to be compatible with current Megatron BERT models, demonstrating expertise in model integration, and framework compatibility.
asrspeech-recognitionnatural-language-processingttsspeaker-diarization
NVIDIA/fsi-samples
Jun 2019 - Oct 2021
A collection of open-source GPU accelerated Python tools and examples for quantitative analyst tasks and leverages RAPIDS AI project, Numba, cuDF, and Dask.
Contributions:8 releases, 67 reviews, 117 commits in 2 years 4 months
cudaanalystpythoncudfleverages
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.