Mirco Ravanelli
Assistant Professor at Mila - Institut Québécois d'Intelligence Artificielle
Montreal, Quebec, Canada
Summary
🤩
Rockstar🎓
Top SchoolMirco Ravanelli is an Assistant Professor at Concordia University and Mila Associate Member specializing in deep learning for Conversational AI and robust sequence processing. With a PhD (cum laude) from the University of Trento and a postdoc at MILA, he has a decade of experience advancing distant-speech recognition and neural architectures for raw-audio processing. He combines academic leadership with hands-on open-source engineering—contributing to widely used toolkits like SpeechBrain and authoring SincNet and pytorch-kaldi enhancements that improve audio front-ends and model stability. His work spans practical improvements (bug fixes, data handling, CTC output filtering) to research innovations in recurrent and cooperative networks for adverse acoustic conditions. Comfortable moving between labs (ICSI, FBK, MILA) and production-oriented codebases, he brings a rare blend of rigorous research pedigree and meticulous engineering craftsmanship.
10 years of coding experience
10 years of employment as a software developer
Doctor of Philosophy - PhD Information and communication technology, Doctor of Philosophy - PhD Information and communication technology at Università di TrentoPost Doctorate Deep Learning for Conversational AI, Post Doctorate Deep Learning for Conversational AI at Université de Montréal
Italian, English, French
Github Skills (22)
pytorch10
python10
data-science10
machine-learning10
audio-processing10
mask-rcnn10
deep-learning10
speaker-recognition10
data-processing10
faster-rcnn10
speech-recognition10
speech-processing9
develop9
recurrent-neural-networks9
asr9
Programming languages (3)
Github contributions (5)
mravanelli/SincNet
Jul 2018 - Mar 2021
SincNet is a neural architecture for efficiently processing raw audio samples.
Role in this project:
ML Engineer
ML EngineerContributions:65 commits, 4 PRs, 57 pushes in 2 years 8 months
Contributions summary:Mirco primarily contributed to a machine-learning project focused on speaker recognition using a SincNet architecture. The commits include updates to the core model definition in `dnn_models.py`, suggesting modifications to the SincNet layers. Further contributions include scripts for computing d-vectors, indicating the user's involvement in feature extraction and potentially downstream tasks like speaker identification. The user also updated data loading and preparation scripts.
audio-samplesmfccspectrogrammlsneural-architecture
mravanelli/pytorch-kaldi
Nov 2018 - Mar 2021
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:68 commits, 31 PRs, 103 pushes in 2 years 4 months
Contributions summary:Mirco primarily contributed to the `pytorch-kaldi` project by modifying and improving the neural network components, specifically the SincConv layer. They addressed issues, improved stability, and fixed compatibility problems with different PyTorch versions. The user also added support for the PASE model within the framework, demonstrating a focus on model development and integration within the speech recognition system.
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.