AI Research Scientist - Clinical NLP

About This Role

The majority of actionable health information is locked in free-text clinical notes, radiology reports, and discharge summaries. Tile Health is building NLP systems that unlock this data to power clinical decision support, risk prediction, and quality measurement. As an AI Research Scientist focused on Clinical NLP, you will design, train, and evaluate transformer-based models tailored to the complexities of medical language, abbreviations, and context.

What You'll Do

• Research and develop novel NLP architectures for clinical entity extraction, relation extraction, and temporal reasoning over medical text
• Curate and annotate clinical text corpora in partnership with physician annotators, ensuring high inter-rater reliability
• Fine-tune large language models on de-identified clinical datasets while maintaining strict HIPAA compliance safeguards
• Design evaluation frameworks that measure model performance across diverse patient populations and clinical specialties
• Publish findings at top-tier venues (ACL, EMNLP, NAACL, CHIL) and contribute to the broader clinical NLP research community
• Collaborate with product engineers to transition research prototypes into production-grade inference services

What We're Looking For

• PhD in Computer Science, Computational Linguistics, or Biomedical Informatics with a focus on NLP
• 3+ publications in peer-reviewed NLP or clinical informatics conferences
• Deep hands-on experience with transformer architectures (BERT, GPT, T5) and frameworks such as Hugging Face Transformers and PyTorch
• Familiarity with clinical text standards and challenges including medical abbreviations, negation detection, and section segmentation
• Experience working with protected health information under IRB or equivalent data governance protocols

Nice to Have

• Prior experience with clinical ontologies (UMLS, SNOMED CT) and biomedical knowledge graphs
• Track record of deploying NLP models into production healthcare systems
• Familiarity with differential privacy or federated learning techniques for sensitive health data
• Experience with active learning or human-in-the-loop annotation workflows