Mixail Petrochuk
Founder & CTO at WellSaid
New York, New York, United States
Summary
🤩
Rockstar🎓
Top SchoolMixail Petrochuk is a founder and CTO with a decade of deep expertise in generative audio, having researched, patented, and productionized models that power real-world products. He co-founded WellSaid in 2018 and scaled it to serve 1,000+ enterprise customers—including 50% of the Fortune 500—while growing ARR to $15M and leading a 30-person R&D organization. Mixail’s work includes industry-standard inventions such as the STFT-based discriminator and patented pronunciations and vocoder breakthroughs that enabled human-parity, long-form, and cost-optimized TTS at scale. He pairs rigorous research (oral presentation at EMNLP 2018 and early open-source impact with PyTorch-NLP) with product-minded engineering that cut inference costs 1,000x and maintained enterprise reliability. A neurodivergent immigrant, he also pioneered ethical guardrails for generative media—withdrawable consent, revenue sharing, moderation, and privacy—reflecting a strong commitment to responsible deployment. Based in Brooklyn, he connects academic breakthroughs to pragmatic systems engineering and still enjoys mentoring and collaborating with local builders.
10 years of coding experience
4 years of employment as a software developer
Computer Science Major and Math Minor, Computer Science, Computer Science Major and Math Minor, Computer Science at University of Washington
Stackoverflow
487reputation
567kreached
7answers
13questions
Github Skills (17)
pytorch10
testing10
machine-learning10
nlp10
metric10
data-science9
python9
datasets8
neural-network8
ssh6
google-cloud-language6
windows6
bash6
pyaudio6
serialization6
Programming languages (11)
Github contributions (5)
PetrochukM/PyTorch-NLP
Feb 2018 - Jul 2021
Basic Utilities for PyTorch Natural Language Processing (NLP)
Role in this project:
Data Scientist & ML Engineer
Data Scientist & ML EngineerContributions:3 releases, 1 review, 384 commits in 3 years 5 months
Contributions summary:Mixail primarily contributed to testing and implementing new metrics for evaluating machine learning models within the PyTorch-NLP library. Their work included writing test cases for accuracy metrics and token-level accuracy, demonstrating a focus on model evaluation. The user also added comprehensive testing for dataset functionality, specifically for the `SimpleQA` and likely other datasets, to ensure correct data loading and processing. This suggests a strong focus on quality assurance for NLP projects within the repository.
pythonbertlanguage-processingword-vectorsnatural-language-processing
PetrochukM/HParams
Jul 2019 - May 2022
A thoughtful approach to hyperparameter management.
Contributions:239 commits, 6 PRs, 203 pushes in 2 years 9 months
pytorchpythondata-sciencedeep-learningapproach
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.