Job Description
📋 Description Own audio side of multimodal generation end-to-end. Design/train audio VAEs, neural codecs, vocoders. Build diffusion/flow-matching models for AV generation. Design audio conditioning and cross-modal alignment in AV models. Run experiments and evaluations; drive data quality and metrics. Lead small projects and collaborate cross-team. 🎯 Requirements Experience with large-scale audio models (>8B params). Diffusion/flow-matching transformers; samplers, schedules, conditioning, distillation. Training audio VAEs, neural codecs, vocoders; objectives. Multi-node, multi-GPU distributed training (FSDP/DeepSpeed). Strong software engineering; production-grade PyTorch code. Shipped speech/audio or multimodal models to production. 🎁 Benefits Competitive salary and equity. Medical, dental, and vision insurance premiums covered. 42 days paid time off annually. 401(k) retirement plan. Lifestyle spending account – $500/month. In-office meals and One Medical membership, and more.
