🛠 Tech Stack

💼 About This Role

You'll join a team of 40+ researchers working on generative speech and voice synthesis to bring in-house voice models to production quality. Your work will directly impact solutions used by over 60,000 businesses worldwide. You'll tackle challenging post-training optimization to achieve real-time, expressive synthetic voices.

🎯 What You'll Do

• Adapt models for new conditioning inputs like emotion, speed, and prosody.
• Fine-tune speech models using DPO, LoRA, and parameter-efficient methods.
• Implement post-training optimization techniques (quantization, pruning, distillation).
• Design and implement new evaluation metrics for TTS systems.

📋 Requirements

• Strong understanding of generative modeling applied to sequential or multimodal data.
• Hands-on experience with LLMs or similar transformer-based architectures.
• High proficiency in PyTorch with distributed training and model optimization.
• Proven experience training deep learning models end-to-end from data to evaluation.

✨ Nice to Have

• Familiarity with state-of-the-art audio generation architectures (diffusion, neural codecs).
• Experience with speech-to-speech or text-to-speech systems.
• Original research contributions at ICASSP, Interspeech, NeurIPS, or ICML.

🎁 Benefits & Perks

• 💰 Competitive compensation with salary, stock options, and bonus.
• 🏠 Fully remote from Europe or hybrid in London, Amsterdam, Zurich, Munich.
• 🗓️ 25 days annual leave plus public holidays.
• 🤝 Great company culture with regular planning and socials at hubs.