Machine Learning Engineer - Voice Conversion

Stack

Responsibilities

• ▹Architect, implement, pre-train, fine-tune and post-train/align (e.g. GRPO/DPO) large-scale speech models
• ▹Design, run and analyze scientific experiments to advance model understanding
• ▹Develop and improve dev tooling to enhance team productivity
• ▹Contribute across the full stack, from low-level optimizations to high-level model design
• ▹Define data requirements and collaborate on acquisition, curation, augmentation, labeling quality and synthetic data strategies
• ▹Design automated objective/subjective evaluations: listening tests, SV/WER/ASR-based metrics, robustness and bias checks, red-team studies
• ▹Harden the training-evaluation-inference pipeline; profile latency, memory and cost; meet production SLAs with monitoring and rollback
• ▹Contribute to safety/consent guardrails and misuse mitigation

Requirements

• ▹Exceptional research/development experience with large-scale audio models (>8B parameters, >500k hours of data)
• ▹Deep hands-on experience with diffusion and/or flow-matching transformers, samplers, schedules, conditioning mechanisms, and distillation
• ▹Deep hands-on experience training audio VAEs, neural audio codecs and vocoders, latent/tokenizer design, reconstruction and perceptual objectives, adversarial training
• ▹Strong experience with multi-node, multi-GPU distributed training (FSDP/DeepSpeed or equivalent)
• ▹Strong software engineering skills with a track record of building complex systems
• ▹Strong PyTorch skills and performance work (profiling, CUDA/Triton/C++ as needed), writing reliable production-quality code
• ▹Experience shipping large-scale speech/audio or multimodal generative models to production

Soft skills

About the company

Cantina is a social platform founded by Sean Parker with the most advanced AI character creator, building lifelike social bots that interact across voice, video and text.