Machine Learning Engineer, Speech - Joint Audio-Video Modeling

Join Cantina Labs to build state-of-the-art speech and audio generation systems for joint audio-video modeling. Own audio representations (audio VAEs, neural codecs, vocoders), generative backbones (diffusion/flow-matching transformers), conditioning/alignment for voice cloning and multi-speaker scenarios, and adjacent tasks (controllable TTS, voice conversion). Lead model/data/eval cycles, design experiments and listening tests, drive inference efficiency (distillation, quantization), productionize training→evaluation→inference pipelines, and partner on GPU scaling and safety/consent guardrails. Requires strong large-scale audio model experience, distributed training, PyTorch and performance/CUDA skills, and production ML engineering.