Research Scientist, Medical World Models

This is a fully remote position, open to applicants in United States, +1 more country.

📋 Description

• Create and train models that predict a member’s future health status based on irregularly sampled histories, incorporating calibrated uncertainty at clinically relevant timeframes.

• Develop encoders that consolidate medical data into integrated representations that are resilient to missing modalities and can be adapted for downstream clinical applications.

• Construct and uphold the evaluation framework, which includes validation sets, confidence intervals, label-efficiency curves, forecasting metrics, collapse diagnostics, and benchmark registration.

• Manage the training and evaluation codebase from start to finish, encompassing data loaders, distributed GPU training, experiment tracking, versioned releases, model cards, and license inventories.

• Provide pretrained encoders and forecasters as well-documented and reproducible artifacts to product-track teams, facilitating clinical validation and transition to Engineering and Product Development.

• Collaborate with the Data Team on dataset specifications and requirements.

• Design analyses that differentiate between predictive and causal claims and make limitations clear.

• Handle de-identified data in a PHI-sensitive, regulated environment and document efforts for regulatory review.

• Maintain experiment logs, design notes, and retrospectives; publish research in leading ML and medical imaging conferences.

• Work alongside ML engineers, Data, Infrastructure, Quality and Regulatory Affairs teams, and clinicians.

• Assist in establishing research reviews, documentation standards, code review processes, and hiring practices.

⛳️ Requirements

• PhD in machine learning, computer science, biomedical engineering, or a related field with 1-2+ years of professional experience, or MS/BS with 5+ years of experience in building and evaluating ML models using real data.

• Proven expertise in at least one of the following areas: temporal/longitudinal modeling (sequence models, neural ODE/CDE or state-space models, forecasting with irregular sampling, survival or progression modeling); self-supervised or foundation-model pretraining on medical imaging (3D MRI/CT) or multimodal data; multimodal integration of imaging with tabular, EHR, or biomarker data.

• Proficient in Python and PyTorch.

• Experience with large-scale training (multi-GPU, extensive datasets, experiment tracking) and writing code that others can utilize.

• Strong evaluation instincts: statistical reasoning, calibration, error analysis, and a consistent practice of questioning whether a result would hold up under a larger validation set.

• A record of publications in top ML or medical imaging conferences (e.g., MICCAI, NeurIPS, ICML, ICLR, CVPR), or equivalent evidence of research contributions that have been implemented in production.

• Effective written and verbal communication skills with both technical and clinical audiences.

🏝️ Benefits

• Stock options.

• Comprehensive health, dental, and vision plans for you and your family.

• Wellness and commuter benefits.

• Competitive vacation policy.

• Remote work flexibility.

• A culture that prioritizes learning, collaboration, and thoughtful engineering.