About the role
Cantina is hiring a Research / ML Engineer for its Speech Team to develop voice conversion and related speech systems, including controllable text-to-speech and voice design. The role spans the full model lifecycle, from data specifications and experimentation through training, evaluation, and production inference. You would work with research, data, infrastructure, and product teams to improve model quality while managing latency, memory, and cost.
The work combines large-scale generative modeling with practical engineering and safety responsibilities. It includes defining and curating training data, building listening tests and objective metrics, hardening pipelines, and contributing to consent and misuse safeguards. The posting seeks substantial hands-on experience with large audio models, diffusion or flow-matching transformers, audio codecs or vocoders, distributed training, and production speech systems.
Highlights
- Build and ship large-scale speech and voice conversion models.
- Design experiments and subjective and automated evaluations.
- Own data quality and training-to-inference pipelines.
- Work on speech safety, consent, and misuse mitigation.
This recap is dataskew's editorial summary, not the company's copy.