Senior/Staff Applied Scientist, Multimodal Representation Learning (Oncology)

Pathos seeks a specialized scientist to develop their Oncology Foundation Model (OFM) stack. The role focuses on designing multimodal AI systems that integrate clinical text, genomics, imaging, and molecular data to improve drug development decisions.

Key Responsibilities

• Design and implement multimodal pretraining strategies for oncology data
• Build cross-modality alignment between clinical narratives and molecular signals
• Create evaluation frameworks beyond standard metrics (ablations, cohort-shift testing, temporal generalization)
• Partner with engineering on scalable GPU training infrastructure
• Package outputs for internal science teams with uncertainty estimates and interpretability

Required Qualifications

• PhD in ML/AI, CS, Statistics, Computational Biology, or related field
• Deep PyTorch experience with large model training
• Demonstrated ability to design and rigorously evaluate representation learning approaches
• Comfort with ambiguous problems and iterative execution

Strongly Preferred

• Multimodal foundation model experience
• Domain expertise in clinical EHR, molecular/omics modeling, or pathology imaging
• Understanding of censoring, batch effects, and confounding in biomedical data

Nice to Have

• Distributed training experience (FSDP/DeepSpeed)
• Alignment methods and robustness evaluation
• Publications at NeurIPS/ICML/ICLR/MLHC or impactful open-source work