I am a Research Engineer in Speech working within the Hi! PARIS / École Polytechnique ecosystem, contributing to research and engineering efforts in speech generation, prosody control, voice conversion, and NLP.
My work combines scientific rigor and practical implementation, with experience spanning:
- French TTS and prosody control
- SSML-based modeling
- WavLM-based speech resynthesis
- evaluation and reproducible research pipelines
- HPC-scale experimentation
I have developed my work in a strong academic research environment, including collaboration and scientific interactions within the École Polytechnique ecosystem.
- 🔊 Controllable French TTS with explicit prosody planning
- 🧠 SSML-based modeling for pauses, rhythm, emphasis, and timing
- 🧪 WavLM → Audio resynthesis with adversarial training and layer ablations
- 🚀 Zero-shot voice conversion using learned speech representations
- 📊 Objective and perceptual evaluation for reproducible speech research
- ⚙️ Distributed training pipelines with PyTorch, DDP, AMP, and Slurm
- 📄 ICNLSP 2025 — Improving French Synthetic Speech Quality via SSML Prosody Control
- 🤗 Released two Hugging Face models for French SSML pause prediction and break rendering
- 🛠️ Built reproducible research pipelines for:
- distributed training (DDP / AMP)
- checkpoint-based evaluation
- ablation studies
- paper-ready tables and figures
- 🎙️ Currently developing WavLM-based speech resynthesis and zero-shot voice conversion pipelines for high-fidelity generation
Hi! PARIS / École Polytechnique
I contribute to research and development in speech generation and language technologies, with a particular focus on:
- controllable TTS
- prosody-aware modeling
- voice conversion
- evaluation methodology
- reproducible experimentation at scale
My work has been carried out in a highly demanding research setting, including scientific interactions with senior researchers such as Éric Moulines.
Controllable French speech synthesis with explicit SSML planning for pauses, timing, and emphasis.
What this project includes
- prosody-oriented text preprocessing
- symbolic pause planning
- SSML generation
- break prediction
- evaluation utilities
- reproducible training and inference scripts
Links
Waveform resynthesis from WavLM representations for speech generation and voice conversion research.
What this project includes
- adversarial waveform reconstruction
- chunked inference with overlap-add
- checkpoint evaluation
- layer ablation experiments
- experiment tracking for paper-ready analysis
Links
A research direction toward zero-shot voice conversion using WavLM-based representations and high-fidelity generation modules.
Current scope
- Stage 1: representation-to-audio resynthesis
- Stage 2: voice conversion pipeline design
- planned diffusion / flow-based conditioning for zero-shot conversion
Links
- ICNLSP 2025 — Improving French Synthetic Speech Quality via SSML Prosody Control
- training and evaluation pipelines
- reproducible experiment configurations
- analysis scripts for ablations
- paper-ready figures and tables
- Text-to-Speech (TTS)
- prosody modeling
- SSML control
- pause prediction
- voice conversion
- speech evaluation
- segmentation and alignment
- chunked inference and overlap-add reconstruction
- PyTorch
- distributed training (DDP)
- mixed precision (AMP)
- GAN training
- speech representation learning
- diffusion / flow-based generation concepts
- Slurm / HPC workflows
- experiment reproducibility
- checkpoint management
- configuration-driven training
- evaluation pipelines
- LaTeX-ready reporting and analysis
| Degree | Institution |
|---|---|
| Master’s degree | Université Gustave Eiffel |
| Master’s degree | UVSQ (Versailles Saint-Quentin-en-Yvelines) |
| Bachelor’s degree | Paris Descartes University |
I am open to research collaborations and industry partnerships in:
- controllable TTS
- prosody modeling
- French speech technology
- voice conversion
- evaluation and reproducibility for speech systems
I am especially interested in projects where scientific rigor, data confidentiality, and engineering quality matter.
If you find my repositories useful for research or development, consider giving them a star on GitHub. It helps increase visibility and supports continued maintenance and improvement.