Yingzhi Wang
Senior Researcher (Speech AI) at Elm Company
Paris, Ile-de-France
Summary
🤩
Rockstar🎓
Top SchoolYingzhi Wang is a Senior Researcher in Speech AI based in Paris with five years of experience building end-to-end speech systems across ASR, TTS, emotion recognition, speaker verification and SLU. Trained at CentraleSupélec with advanced study in optimization and robotics, she transitioned from industrial research roles at Zaion and Orange to lead speech AI work at Elm Company. A core collaborator on the widely used SpeechBrain toolkit, her open-source contributions focus on integrating Wav2Vec2 for emotion recognition and refining training recipes and data handling for IEMOCAP. She combines hands-on ML engineering with research rigor, iterating on architectures, loss functions and learning schedules to push practical model performance. Known for bridging production needs and novel speech LLM experiments, she often explores non-obvious design choices like average time pooling and speaker-aware inputs. Her profile reflects a pragmatic researcher who ships reproducible tools and advances applied speech science.
5 years of coding experience
5 years of employment as a software developer
Engineer's degree Interactive systems and robotics, Engineer's degree Interactive systems and robotics at CentraleSupélecMaster's degree Optimization & Algorithms, Master's degree Optimization & Algorithms at Universite de LorraineBachelor's degree Automatics, Bachelor's degree Automatics at Dalian University of Technology
Github Skills (7)
pytorch10
machine-learning10
audio-processing10
deep-learning10
speech-processing10
emotion-recognition10
wafv210
Programming languages (2)
Github contributions (5)
speechbrain/speechbrain
Aug 2021 - Dec 2022
A PyTorch-based Speech Toolkit
Role in this project:
ML Engineer
ML EngineerContributions:40 reviews, 157 commits, 23 PRs in 1 year 3 months
Contributions summary:Yingzhi's commits primarily focus on integrating and experimenting with the Wav2Vec2 model for emotion recognition within the speechbrain framework. They implemented and modified a recipe for training an emotion recognition system using the IEMOCAP dataset and utilized a pre-trained Wav2Vec2 model, exploring variations in the architecture like average time pooling and the use of different loss functions. The commits also involved updates to the training process, including modifications to learning rate schedules and the handling of data inputs, such as adding a model and speaker.
voice-recognitionasrspeech-recognitionspeech-separationspeaker-verification
BenoitWang/speechbrain
Aug 2021 - Aug 2024
A PyTorch-based Speech Toolkit
Contributions:193 pushes, 36 branches in 2 years 11 months
pytorchspeech-toolkitspeech-to-textspeech-recognitionspeech-synthesis
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.