ML Infrastructure Engineer, Real-time Inference
Make speech models fast and reliable enough for live phone calls, with a target response time under 300 ms.
About the role
You’ll build the serving stack behind the Waridi API: streaming inference, GPU scheduling, telephony integration and the observability to keep it all honest.
What you’ll do
- Build and operate low-latency streaming inference services
- Integrate with telephony over SIP and WebRTC
- Own latency, reliability and cost across the voice pipeline
- Make deployment and evaluation of new models routine
What we’re looking for
- Production experience with GPU inference or real-time systems
- Strong in Python plus one systems language (Go, Rust or C++)
- Kubernetes or equivalent orchestration
Nice to have
- WebRTC or VoIP experience
- Model quantisation and optimisation
- Experience running infrastructure in Africa
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.