Postdoctoral Researcher Foundation Models

• Design, implement, and benchmark foundation-model architectures for microbiome data, including transformer-based and masked-autoencoder approaches and relevant architectures adapted from related biological domains.
• Develop representations that integrate microbial identity and abundance with genomic or functional information and contextual metadata such as crop genotype, soil, and environmental conditions.
• Define and evaluate self-supervised learning objectives and embedding strategies, and benchmark their added value against simpler machine-learning baselines.
• Train and evaluate the AI foundation model on large-scale microbiome datasets, with attention to sparsity, batch effects, scalability, generalization across studies, and uncertainty in downstream predictions.
• Fine-tune the model for tasks including microbial root competence and crop-relevant outcomes, and iteratively improve the model using experimental Design-Build-Test-Learn data generated by project partners.
• Develop a generative component, exploring autoencoder- and/or diffusion-based approaches for generating ecologically plausible microbiome configurations.
• Apply interpretable and explainable AI approaches to identify microbial taxa, functions, and contextual features driving model predictions.
• Develop reproducible training and evaluation workflows and work with project partners to make models and associated tools usable beyond the immediate research setting.

The model's predictions will be tested experimentally in greenhouse and field settings, and the resulting microbiome and crop phenotype data will feed back into model development. This gives you the opportunity to develop new AI methodology while seeing how model predictions perform in a real biological and agricultural system.

In this position, you will be part of an interdisciplinary research environment spanning AI technology for life and plant-microbe interactions groups, with close interaction with bioinformatics, microbial ecology, and experimental crop research at the institution and project partners. You will have access to the institution's GPU/HPC infrastructure and large, curated microbiome and microbial genome datasets.

Additionally, you will collaborate closely with project partners at diverse institutions, including experimental teams that will directly test model predictions.

What they ask

• A PhD, or a PhD close to completion, in machine learning, artificial intelligence, computational biology, bioinformatics, computer science, or a closely related field.
• Strong hands-on experience with deep learning and modern representation learning, preferably including transformers, self-supervised learning, foundation models, autoencoders, or related architectures.
• Strong programming skills in Python and experience with a deep-learning framework such as PyTorch, including training and evaluating models on GPU/HPC infrastructure.
• Experience working with high-dimensional biological, omics, ecological, or similarly sparse and heterogeneous data, or a clear motivation to develop this expertise.
• An interest in interpretable AI, rigorous benchmarking, and reproducible research, together with the ability to collaborate across AI, bioinformatics, microbiology, and crop science.

Wishes.

• Experience with microbiome data, microbial genomics, metagenomics, or multimodal biological data is an advantage, but is not required if you bring strong machine-learning expertise and are motivated to learn the biology.

• A central role in developing the core AI technology of a five-year European research project.
• Opportunities for personal and professional growth, flexible leave arrangements, and extra vacation days.
• You can tailor your employment package to your needs.