Postdoc: Build AI Foundation Models for Crop Microbiomes

Sport Society; Utrecht; On-site; EUR 41,000 - 64,000; Full time

Benefits offered by this job

Holiday pay

Year-end bonus

Pension scheme

Flexible terms

Job summary

Utrecht University is seeking a Postdoctoral Researcher to lead the development of ARCA, a crop microbiome foundation model, within the NOAH project. You will explore architectures and objectives for sparse, high-dimensional data and translate methods into robust trainable models for pretraining and downstream tasks.

Your work spans deep learning, bioinformatics and microbial ecology, with collaboration across UU, Aarhus, Niab, INRAE and The Hyve.

Qualifications

• PhD in machine learning, artificial intelligence, computational biology, bioinformatics, computer science or related field.
• Strong hands-on experience with deep learning and representation learning (transformers, self-supervised, foundation models, autoencoders).
• Strong Python programming skills and experience with PyTorch, including GPU/HPC training/evaluation.
• Experience with high-dimensional biological, omics, ecological or similarly sparse data; motivation to develop this area.
• Interest in interpretable AI, rigorous benchmarking and reproducible research, with collaboration across AI, bioinformatics, microbiology and crop science.

Responsibilities

• Design, implement and benchmark foundation-model architectures for microbiome data, including transformer-based and MAE approaches.
• Develop representations integrating microbial identity, abundance, genomic/functional info and contextual metadata.
• Define and evaluate self-supervised objectives and embedding strategies; benchmark against baselines.
• Train and evaluate ARCA on large-scale microbiome datasets, addressing sparsity, batch effects, scalability and uncertainty.
• Fine-tune ARCA for microbial root competence and crop-relevant outcomes; iteratively improve via Design-Build-Test-Learn cycles.
• Develop generative ARCA components using autoencoder- or diffusion-based approaches.
• Apply interpretable AI to identify microbial taxa, functions and contextual drivers of predictions.
• Develop reproducible training/evaluation workflows and collaborate to deploy tools beyond research.

Skills

Deep learning

Transformers

Self-supervised learning

Python

PyTorch

GPUs / HPC

Interdisciplinary collaboration

Interpretable AI

Education

PhD in ML / computational biology / bioinformatics

Tools

PyTorch

GPUs / HPC

Job description

Utrecht University is seeking a Postdoctoral Researcher to lead the development of ARCA, a crop microbiome foundation model, within the NOAH project. You will explore architectures and objectives for sparse, high-dimensional data and translate methods into robust trainable models for pretraining and downstream tasks.

Your work spans deep learning, bioinformatics and microbial ecology, with collaboration across UU, Aarhus, Niab, INRAE and The Hyve.