Edresson Casanova
Senior Research Scientist at NVIDIA
São Carlos, São Paulo, Brazil
Summary
🤩
Rockstar🎓
Top SchoolEdresson Casanova is a Senior Research Scientist and TTS deep learning engineer with nine years of experience specializing in speech processing, including speech synthesis, ASR, speaker verification, and speech-based illness detection. He holds a PhD from USP and a bachelor's from UTFPR, and has transitioned research prototypes into production-ready tooling at Coqui and now NVIDIA. An active open-source contributor, he made notable engineering contributions to widely used projects like Mozilla TTS and Coqui TTS—adding data pipelines, spectrogram extraction, training/inference improvements and unit tests for robustness. Beyond model work, he combines systems-minded engineering (optimizer and training script tuning) with linguistic-focused research such as zero-shot multi-speaker TTS and low-resource ASR, enabling practical multilingual and clinical applications. Based in São Carlos, Brazil, he pairs deep academic rigor with hands-on implementation that bridges research and deployable speech systems.
9 years of coding experience
2 years of employment as a software developer
PhD. Speech Processing and Natural Language Processing, PhD. Speech Processing and Natural Language Processing at Instituto de Ciências Matemáticas e de Computação (ICMC) - USPBachelor of Computer Science Speech Processing and Natural Language Processing, Bachelor of Computer Science Speech Processing and Natural Language Processing at Federal University of Technology - Parana
Stackoverflow
31reputation
1kreached
1answer
0questions
Github Skills (18)
speech-to-text10
pytorch10
python10
machine-learning10
text-to-speech10
ml10
deep-learning10
data-processing10
configuration-management9
natural-language-processing9
trainings9
modeling9
nlp9
unit-testing8
librosa6
Programming languages (7)
Github contributions (5)
coqui-ai/TTS
Apr 2021 - Jan 2023
🐸💬 - a deep learning toolkit for Text-to-Speech, battle-tested in research and production
Role in this project:
Back-end Developer & ML Engineer
Back-end Developer & ML EngineerContributions:1 release, 179 reviews, 397 commits in 1 year 8 months
Contributions summary:Edresson implemented changes to the optimizer initialization and other parameters in training scripts for the HiFi-GAN vocoder and the VITS model. They also added a script for the extraction of TTS spectrograms, a core component of text-to-speech pipelines. Further contributions included creating a new inference function and adding unit tests for the extraction of spectrograms, indicating a focus on both the model training and inference processes within the TTS framework.
deep-learningtext-to-speechpythonspeechpytorch
mozilla/TTS
Mar 2020 - Oct 2020
:robot: :speech_balloon: Deep learning for Text to Speech (Discussion forum: https://discourse.mozilla.org/c/tts)
Role in this project:ML Engineer
ML EngineerContributions:1 review, 53 commits, 15 PRs in 8 months
Contributions summary:Edresson made significant contributions to the `mozilla/tts` repository, which focuses on deep learning for text-to-speech. Their work centered around adding text parameters to the configuration files and modifying the `TTSDataset.py`, `utils/text/__init__.py`, and `utils/generic_utils.py` files, suggesting involvement in data processing, text-to-sequence conversion, and model configuration, all essential components of a text-to-speech system. The user also updated the `notebooks/ExtractTTSpectrogram.ipynb`, `train.py`, `utils/text/symbols.py`, and `synthesize.py` files, to improve the models training and usability.
deep-learningtext-to-speechpythonpytorchtacotron
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.