Research Staff, Voice AI Foundations [UK - Remote]
Deepgram · Remote · London, UK ·
COMPENSATION
€•• – ••k
Listed on the employer's job board (Ashby)
- Fully remote (UK)
- Voice AI Research
- Latent Space Models
- Senior Research Role
- AI-first Culture
Join Deepgram's research team to solve fundamental data and scale challenges in voice interaction. You will pioneer Latent Space Models, focusing on extreme compression, high-fidelity reconstruction, and steerable generative models for human speech.
The company expects an AI-first mindset, requiring active use of advanced tools to innovate and measure performance. This role suits researchers who thrive in rapid change and enjoy experimenting with new models daily.
What you'll do
- Build next-gen neural audio codecs for low bit-rate compression and high fidelity
- Pioneer steerable generative models to synthesize diverse human speech styles
- Develop embedding systems that factorize codec latent space into interpretable dimensions
- Generate synthetic audio data at scale using latent recombination techniques
- Design model architectures and training schemes optimized for hardware efficiency
What you'll need
- Strong mathematical foundation in statistical learning theory and self-supervised learning
- Deep expertise in foundation model architectures and multimodal scaling
- Proven ability to bridge theory with efficient practical implementation
- Experience building data pipelines for massive dataset processing and curation
- Track record of designing controlled experiments to validate architectural innovations
- Knowledge of hardware constraints and techniques for real-world model optimization
- History of open-source contributions or publications advancing speech/language AI
Stack
- Python
- PyTorch
- Triton
- TPU
- VQ-VAE
- Mamba-2
How you'll work
- Fully remote position
- Based in the United Kingdom
- Fast-paced environment with rapid iteration
- Requires active daily use of advanced AI tools