Research Scientist, Speech Recognition
Train Waridi Sikia to hear Swahili the way people actually speak it — accents, code-switching and phone-quality audio included.
About the role
You’ll own research on dialect-aware speech recognition: from data mixtures and training recipes to evaluation on our Swahili benchmark. Your work ships directly into Waridi Sikia and the voice agents built on it.
What you’ll do
- Design and run training experiments for streaming and batch ASR models
- Improve robustness to regional accents, noise and Swahili–English code-switching
- Help build and maintain our Swahili ASR benchmark
- Write up what you learn for the team and, where we can, for the public
What we’re looking for
- Hands-on experience training speech or sequence models in PyTorch or JAX
- Strong fundamentals in machine learning and evaluation
- Comfort working from messy, real-world audio data
Nice to have
- Native or fluent Swahili
- Experience with low-resource or multilingual speech
- Published research in speech or NLP
How we hire
A short intro call, a practical exercise close to the real work, a conversation with the team, and a decision within two weeks. We don’t expect you to tick every box — if this excites you, apply.