Senior Applied Scientist, Real-Time Multimodal AI

Amazon; Seattle (WA); On-site; USD 167,000 - 226,000; Full time

Job summary

Amazon is seeking a Senior Applied Scientist to drive real-time multimodal conversational AI research and development. You will own a research area across pre-training to deployment, scaling foundation models for speech and audio, and building post-training systems such as reward modeling and RL alignment.

You will collaborate across scientists and engineers to ensure models are production-ready, with attention to latency, data efficiency, and real-time performance.

Qualifications

• 5+ years building ML models for business applications.
• PhD or MS with 6+ years of applied research experience.
• Experience programming in Java, C++, Python or related language.
• Hands-on experience training large-scale foundation models.
• Experience with multimodal architectures across speech, text, and audio.
• Strong understanding of transformer architectures for speech/audio.
• Publication record at top-tier venues.
• Demonstrated ownership of a research area and cross-team collaboration.

Responsibilities

• Scale foundation models for real-time speech and audio generation.
• Design architectures considering hardware constraints for inference.
• Develop training methodologies for multimodal models.
• Contribute to post-training methods like RL and reward modeling.
• Collaborate with engineers to deploy models for real-time use.

Skills

Large-scale ML training

Multimodal models

Transformer models

Research leadership

Publications

Education

PhD

MS + 6+ years

Tools

Python

Java

C++

Job description

Amazon is seeking a Senior Applied Scientist to drive real-time multimodal conversational AI research and development. You will own a research area across pre-training to deployment, scaling foundation models for speech and audio, and building post-training systems such as reward modeling and RL alignment.

You will collaborate across scientists and engineers to ensure models are production-ready, with attention to latency, data efficiency, and real-time performance.