Vitaly Lavrukhin
Senior Research Manager at NVIDIA
California, United States
Summary
🤩
Rockstar🎓
Top SchoolVitaly Lavrukhin is a Senior Research Manager at NVIDIA with 12 years of experience building and managing teams focused on Speech AI and conversational systems. He leads applied research on NeMo, contributing hands-on to ASR model improvements, streaming demos, and tooling that help move large speech models from research to production. His background spans academia and industry—from supervising AI research as a university professor to engineering DSP/FPGA implementations and leading deep learning efforts at Samsung R&D. Known for pragmatic engineering, he blends algorithmic depth (PhD-level computer science) with production-oriented DevOps and backend work, including contributions to widely used open-source toolkits like NVIDIA/NeMo and OpenSeq2Seq. Based in California, he drives cross-functional workflows for data processing and model deployment at scale. Colleagues describe him as a research leader who still dives into code and scripts to unblock teams and ship demonstrable features.
12 years of coding experience
16 years of employment as a software developer
CourseraPhD, Computer Science, PhD, Computer Science at Moscow Power Engineering Institute (Technical University)
Russian, English
Github Skills (19)
pytorch10
python10
scripting10
machine-learning10
script10
deep-learning10
tensorflow10
asr10
sh10
speech-recognition10
shell10
language-modeling9
trainings9
configuration-file9
configurations9
Programming languages (3)
Github contributions (5)
NVIDIA/OpenSeq2Seq
Apr 2018 - Apr 2021
Toolkit for efficient experimentation with Speech Recognition, Text2Speech and NLP
Role in this project:
Back-end Developer & DevOps Engineer
Back-end Developer & DevOps EngineerContributions:3 releases, 1 review, 241 commits in 3 years
Contributions summary:Vitaly primarily contributed to the speech recognition toolkit by updating and fixing the download scripts, documentation, and model configuration files. Their work involved modifying shell scripts for downloading language models and adapting documentation to reflect the latest updates. They also made changes to the configurations of various speech-to-text models and tests, indicating a focus on improving the functionality and usability of the provided models within the toolkit. Additionally, the user performed modifications in the custom CTC beam search decoder and incorporated changes to support different pre- and post-normalization schemes, enhancing the model's performance.
nlpexperimentationseq2seqspeech-to-textrecognition
NVIDIA/NeMo
Sep 2019 - 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:75 reviews, 41 commits, 42 PRs in 3 years 4 months
Contributions summary:Vitaly contributed to the development and improvement of the NeMo framework, specifically focusing on Automatic Speech Recognition (ASR) models. Their work included adding and optimizing features like checkpoint loading, saving log probabilities for evaluation, and supporting fixed normalization techniques. They also added examples, including an offline ASR inference notebook and a streaming ASR demo.
asrspeech-recognitionnatural-language-processingttsspeaker-diarization
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.