Research Engineer/Research Scientist, Audio

About this role

Anthropic's Audio team seeks a Research Engineer/Scientist to develop safe, steerable audio and speech systems using large language models. You'll work across the full ML stack—from audio codecs and data synthesis to large-scale model training and production deployment—in a collaborative environment balancing research and engineering.

What you'll do

• Develop audio codecs, representations, and data synthesis pipelines for high-quality audio
• Train and optimize large-scale speech language models and audio diffusion models
• Design novel architectures for integrating continuous audio signals into LLMs
• Build conversational systems, speech understanding models, and speech synthesis capabilities
• Collaborate with teams across pretraining, finetuning, and production inference to deploy audio technologies
• Create evaluation methodologies for qualities like naturalness, expressiveness, and robustness

What they're looking for

• Audio model training (speech-to-speech, speech recognition, text-to-speech, codecs, generative audio)
• PyTorch or JAX with distributed training expertise
• Signal processing and audio fundamentals
• Large-scale model training and inference optimization
• Diffusion models for audio generation
• Reinforcement learning for language models
• GPU and distributed training infrastructure (Kubernetes, etc.)
• System-level debugging and performance benchmarking

Benefits

• Competitive annual salary range: $350,000–$500,000 USD
• Hybrid work policy with minimum 25% office presence in San Francisco
• Visa sponsorship available
• Opportunity to shape responsible voice AI development
• Collaborative environment across company-wide teams
• Work on cutting-edge audio AI research with real-world impact