Research Scientist, Frontier Health, DeepMind

Job description

Conduct fundamental and applied ML research to develop physiological and behavioral world models simulating continuous-time human biology. Design novel machine learning architectures (e.g., state-space models, neural dynamical systems) for multi-modal clinical telemetry and longitudinal EHRs (e.g., MIMIC-IV). Build and maintain robust evaluation benchmarks (OxyBench) to assess disease trajectory predictions, acute clinical events (e.g., sepsis), and counterfactual treatment simulations. Collaborate cross-functionally with ML researchers, software engineers, and external clinical partners across Mountain View, London, and Paris. Publish original research in top machine learning conferences and leading medical journals. Minimum Qualifications: PhD in Computer Science, Machine Learning, Computational Biology, Applied Mathematics, Physics, or equivalent practical experience. 2 years of experience (industry or internships) in building world models for adaptive systems, foundation models, continuous dynamical systems, state-space models, and deep generative architectures. Experience with model robustness, out-of-distribution generalization, and uncertainty quantification. Research experience with first-author publications at machine learning venues or domain journals (NeurIPS, ICML, ICLR, etc.). Preferred Qualifications: Strong experience learning underlying system dynamics from partially observable environments, managing irregular sampling, missing modalities, and latent state estimation. Experience applying these methodologies to biomedical domains, framing multi-modal healthcare data, longitudinal EHRs (e.g., MIMIC-IV), and physiological telemetry as complex adaptive systems. Proficiency in Python and modern deep learning frameworks (JAX, PyTorch, or TensorFlow). Strong collaborative skills for working in interdisciplinary teams alongside clinical partners. Passion for AI technology and all of its possibilities.