Titouan Parcollet
Senior Research Scientist at Samsung Electronics
Cambridge, England, United Kingdom
Summary
🤩
Rockstar🎓
Top SchoolTitouan Parcollet is a Senior Research Scientist with 11 years of experience building efficient, scalable speech and self-supervised learning systems, currently based in Cambridge and co-leading the open-source SpeechBrain toolkit. He has accelerated large-scale model training and halved VRAM requirements for production speech recognizers, and holds patents for linear alternatives to self-attention tailored to streaming and on-device use. As an academic practitioner he is an Affiliated Lecturer at the University of Cambridge and an associate professor on leave from Avignon Université, bridging rigorous research with practical deployment. His contributions to widely used projects like pytorch-kaldi and SpeechBrain demonstrate deep expertise in audio processing pipelines, model architectures, and decoding systems. Notably, he developed methods to adapt on-device recognizers with only one minute of unlabeled data, achieving substantial WER reductions without supervision. He combines a PhD in computer science with a track record of shipping production-ready, resource-efficient AI for real-world and on-device scenarios.
11 years of coding experience
9 years of employment as a software developer
Doctor of Philosophy - PhD, Computer Science, Doctor of Philosophy - PhD, Computer Science at Avignon Université
fancais, English, Italian, Spanish
Stackoverflow
27reputation
2kreached
0answers
4questions
Github Skills (20)
pytorch10
python10
rnn-model10
n10
kaldi10
deep-learning10
speech-recognition10
machine-learning9
recurrent-neural-networks9
asr9
deep-neural-networks9
neural-network9
documentation9
configuration-management8
cuda6
Programming languages (3)
Github contributions (5)
speechbrain/speechbrain
Mar 2020 - Jan 2023
A PyTorch-based Speech Toolkit
Role in this project:
Back-end Developer
Back-end DeveloperContributions:1 release, 311 reviews, 1013 commits in 2 years 10 months
Contributions summary:Titouan's contributions focused on implementing pooling and architecture modifications within the SpeechBrain toolkit. This involves adding new functionalities to the nnet/architectures.py file, indicating work with the core neural network components. The user's edits include detailed documentation and example code, which is beneficial for developers.
voice-recognitionasrspeech-recognitionspeech-separationspeaker-verification
mravanelli/pytorch-kaldi
Nov 2018 - May 2020
pytorch-kaldi is a project for developing state-of-the-art DNN/RNN hybrid speech recognition systems. The DNN part is managed by pytorch, while feature extraction, label computation, and decoding are performed with the kaldi toolkit.
Role in this project:ML Engineer
ML EngineerContributions:28 commits, 6 PRs, 24 pushes in 1 year 6 months
Contributions summary:Titouan primarily contributes to the core functionality of the pytorch-kaldi speech recognition system. Their commits focus on integrating and improving the system's ability to process audio, including adding features for decoding custom audio files and optimizing the averaging of loss values. They also made changes related to the learning rate, addressing a bug with learning rate halving. Their work touched upon the system's core configuration, including the production setup and data processing pipelines.
extractiondevelopingasrctcspeech-recognition
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.