PhD in natural language processing and machine learning for healthcare

You will join the CaRe-NLP: Human-Centric and Responsible NLP methods for Dutch healthcare project. CaRe-NLP's main goal is to develop human-centric and responsible NLP and ML methods for healthcare in the Netherlands, Europe, and worldwide. CaRe-NLP's focus is methodological, meaning you will focus on developing, building, and testing state-of-the-art large language models (LLMs) that are interpretable and explainable, ensure privacy and fairness, prevent bias, and cope with data scarcity. 

This PhD project will be co-supervised by dr. Iacer Calixto and prof. Ameen Abu-Hanna at the University of Amsterdam and together with prof. Barbara Plank from the MaiNLP research lab at LMU Munich. 

You will tackle relevant clinical problems and develop methods for non-English (including Dutch) multi-modal electronic heath records (EHRs) data that include (combinations of) free-text clinical notes collected in primary and/or secondary settings (including intensive care); medical images; time series measurements; medical knowledge graphs. 

We currently offer the PhD position titled PhD-quantifying-disagreement-and-uncertainty where you will be responsible for developing, building and testing clinical LLMs that model uncertainty and disagreement in annotations (input) and predictions (output). You will develop solutions and apply them to learning from patient electronic health records (EHRs) and free text.

Furthermore, you will:

• Liaise with a network of clinical partners of the CaRe-NLP project and will tackle relevant healthcare problems using rich real-world clinical data;
• Have access to excellent HPC facilities and strong support to publish the results of your work in the best scientific conferences and journals in NLP, ML, (medical) AI, and healthcare in general;
• Contribute to the mentoring of master's students on topics related to your research.

You have:

• A MSc degree in computer science, natural language processing, machine learning, artificial intelligence, statistics, or a related discipline;
• A strong technical background and experience with natural language processing and machine learning methods, toolkits, and best practices;
• Experience with or interest in learning about pretraining and adapting LLMs using large datasets and/or uncertainty quantification with neural networks and generative models (e.g. LLMs, VAEs, GANs, diffusion models);
• Programming skills in Python and ML libraries (PyTorch);
• A responsible, independent and resilient mindset, and you speak English fluently.

Furthermore, these are required but helpful:

• You have experience with data processing pipelines and with pre-training/fine-tuning/deploying Large Language Models in HPC;
• You contributed to open-source projects;
• You have experience with clinical data and related application areas (e.g., cancer, primary/intensive care);
• You have knowledge of the Dutch language, since much of the data you will work with includes free-text clinical notes in Dutch.