About the role
Cantina is hiring a Machine Learning Engineer, Ops to productionize text-to-speech and speech recognition models. The research team supplies trained checkpoints; this role turns them into low-latency, real-time streaming services and owns the inference stack, Kubernetes infrastructure, and validation needed to release changes safely. It connects model research with production operations for a social AI platform focused on expressive, interactive experiences.
The work spans inference engine optimization, GPU performance, model accuracy regression testing, deployment, CI/CD, observability, and cost measurement. Candidates should have hands-on experience deploying and scaling ML models in production, with Python, PyTorch, GPU profiling, Kubernetes, cloud infrastructure, and model-serving knowledge. Experience with speech or audio models and frameworks such as vLLM, SGLang, TensorRT-LLM, or Triton is relevant.
Highlights
- Optimize streaming TTS and ASR inference for latency and throughput.
- Deploy and operate GPU services on Kubernetes.
- Validate numerical parity and model accuracy across releases.
- Track performance, utilization, and cost per request.
This recap is dataskew's editorial summary, not the company's copy.