Pavel Korshunov
Research Associate
Martigny, Wallis, Switzerland
Summary
👤
Senior🎓
Top SchoolPavel Korshunov is a research associate and machine learning specialist with over a decade of experience in biometrics, deepfake and tampering detection, and speech diarization, grounded in a PhD from the National University of Singapore. Based at Idiap in Switzerland, he combines rigorous, reproducible open‑source research with practical system work in C/C++ and Python, and has authored 60+ publications including multiple best‑paper recognitions. His contributions to widely used projects like pyannote‑audio show hands‑on impact—fixing model loading, improving audio I/O, and stabilizing augmentation pipelines for speaker diarization. He also brings a rare blend of multimedia expertise spanning visual privacy, HDR/UHD imaging and crowdsourced QoE, and has regularly taught biometrics and reproducible research courses. Colleagues describe him as a methodical problem‑solver who bridges academic rigor and production readiness in multimedia AI.
10 years of coding experience
7 years of employment as a software developer
Specialist, Computer Science, Specialist, Computer Science at Saint Petersburg State UniversityPh.D., Computer Science, Ph.D., Computer Science at National University of Singapore
English, Russian, French
Github Skills (7)
speaker-diarization10
pytorch10
speech-processing10
python10
sound-files9
machine-learning9
librosa8
Programming languages (2)
Github contributions (5)
pyannote/pyannote-audio
Nov 2018 - Mar 2019
Neural building blocks for speaker diarization: speech activity detection, speaker change detection, overlapped speech detection, speaker embedding
Role in this project:
ML Engineer
ML EngineerContributions:5 commits, 5 PRs, 29 comments in 4 months
Contributions summary:Pavel primarily contributed to the pyannote-audio repository by fixing bugs related to model loading on CPU, incorporating a subset option across all 'apply' modes, and addressing labeling inconsistencies. They also improved the audio feature extraction process by switching from older libraries to soundfile for better format support and adding channel selection during cropping. Furthermore, the user corrected an issue concerning noise accumulation within the augmentation module.
diarizationspeech-recognitionspeaker-verificationspeech-processingspeaker-diarization
pkorshunov/pyannote-audio
Nov 2018 - Nov 2019
Neural building blocks for speaker diarization: speech activity detection, speaker change detection, speaker embedding
Contributions:20 pushes, 1 branch in 1 year
speaker-diarizationbuilding-blocksdiarizationspeech-activity-detectiondeep-learning
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.