Senior AI Scientist

About the role

The Senior AI Scientist will develop cutting-edge medical models for Atria's clinical AI agenda, focusing on personalized and preventive care using a diverse array of data sources.

Responsibilities

• Own the medical modeling roadmap, identifying and building models that significantly improve clinical outcomes.
• Survey and recommend fine-tuning and distillation strategies for open-source models relevant to Atria's needs.
• Design and validate novel architectures for complex, longitudinal multi-modal data.
• Curate and evaluate high-quality training datasets, ensuring they are representative and free of contamination.
• Run rigorous evaluations, including held-out clinical benchmarks and comparisons against leading closed-source models.
• Stay updated with the latest training and fine-tuning methods, and apply them to Atria's specific problems.
• Collaborate with clinicians and research partners to ensure models meet clinical standards and address real-world challenges.

Requirements

• A graduate degree (PhD preferred) in computer science, machine learning, computational biology, biomedical informatics, or a related field.
• Strong hands-on experience training and fine-tuning modern deep learning models, with a track record of published or deployed models.
• Expertise in modern fine-tuning and post-training methods, including SFT, PEFT, preference tuning, distillation, and continued pre-training.
• Deep knowledge of the open-source model ecosystem, with practical experience in training infrastructure and evaluation practices.
• Genuine interest in healthcare and a commitment to ensuring models positively impact patient care.

Qualifications

• Experience building multimodal medical models, integrating clinical text with imaging, structured labs, or physiological signals.
• Familiarity with clinical/biomedical foundation models and the medical model literature.
• Experience with reinforcement learning, reasoning model training, or other frontier post-training techniques.
• Interest in model interpretability and uncertainty quantification for clinical settings.

Skills

• Python and PyTorch proficiency.
• Hands-on experience with the Hugging Face ecosystem and other training stacks.
• Experience with distributed training, mixed precision, efficient data loading, and experiment tracking tools.
• Ability to design and run rigorous evaluations, including calibration analysis and subgroup performance studies.
• Strong communication skills, especially in collaborating with clinicians and research partners.

Benefits

• Excellent health and wellness benefits.
• OneMedical membership for employees and dependents.
• Preventive health screenings through partner hospitals.
• Wellness perks, including fitness programs and financial incentives.
• 401(k) contributions and a company match.
• Flexible time off and continuing medical education support.