Member of Technical Staff — Pretraining Infra

About the Company

Nuance Labs is building photorealistic, real-time AI avatars with emotional intelligence: a full-duplex audiovisual system that can listen, speak, react, interrupt, and respond like a real person.

Responsibilities

• Own the distributed training stack for omni model pretraining from 0→1 system design to 1→10 scaling
• Build and operate the core training runtime including job orchestration, distributed execution, and checkpointing
• Optimize large-scale training performance across parallelism strategy, GPU communication, and memory usage
• Build infrastructure for omni training workloads including high-throughput audio/video/text data loading
• Evolve the platform as model architectures and research directions change

Requirements

• Hands-on experience running large-scale distributed training jobs across large GPU clusters (100s to 1,000+ GPUs)
• Deep understanding of distributed training mechanics like data/tensor/pipeline/sequence parallelism
• Strong understanding of GPU communication and performance debugging including NCCL and collectives
• Practical experience with Megatron, PyTorch FSDP, DeepSpeed, or equivalent infrastructure
• Understanding of multimodal training challenges such as long temporal context and variable sequence lengths
• Strong software engineering fundamentals

Preferred Qualifications

• Prior 0→1 experience building large-scale training infrastructure
• Experience training large omni or multimodal models involving audio, video, or text
• Experience with RL, post-training, data infrastructure, or serving
• Publications or open-source contributions in ML systems, distributed systems, or HPC

Benefits

• $300,000 – $400,000 base salary plus meaningful equity
• HSA plan with ~$2,000 annual company contribution
• 15 days PTO plus public holidays and a full week off at year-end
• Daily lunch, drinks, and snacks
• Commuter benefits
• Visa sponsorship available (O-1, H-1B, green card)