Principal Applied Scientist

What you'll do

• Develop methods to adapt foundation models for enterprise tasks like document understanding and workflow automation.
• Build scalable training and inference systems for RLHF, fine-tuning, and parameter-efficient adaptation techniques.
• Implement inference-time steering and controllability approaches to improve model behavior.
• Integrate privacy-preserving and security-focused training techniques such as differential privacy into production pipelines.
• Develop tools for evaluating, debugging, and monitoring the performance of foundation models.
• Translate customer requirements into scalable technical solutions for domain specialization and agentic behaviors.
• Drive technical projects from initial concept to prototype to deliver new systems components and empirical insights.

What we're looking for

• Bachelor's degree plus 8 years of experience, Master's plus 6 years, or Doctorate plus 5 years in a relevant technical field.
• Ability to pass Microsoft Cloud background checks and other required security screenings.
• Experience with Python and machine learning frameworks like PyTorch (preferred).
• Experience building scalable ML systems or pipelines for training, adapting, or deploying AI models (preferred).
• Experience contributing to research, open-source systems, or production deployments involving foundation model training or adaptation (preferred).
• Experience in areas such as RLHF, transformer architectures, privacy-preserving ML, or multimodal model training (preferred).
• 7+ years of experience conducting research in academic or industry settings (preferred).
• 5+ years of experience developing and deploying live production systems as part of a product team (preferred).