Postdoctoral Position in Machine Learning for Automated Plant Phenotyping (PhenoMix Project)

The postdoc will develop and implement cutting-edge machine learning approaches for automated trait estimation, focusing on: Foundation Models for Phenotyping: Leveraging and adapting pre-trained foundation models for crop trait estimation in both pure stands and crop mixtures, minimising computational and data annotation overheads while maximising generalisation power Domain Transfer Methods: Developing plant-aware image-based domain transfer techniques to enable models trained on high-resolution FIP images to work effectively with lean device images (e.g., smartphone cameras) 3D Reconstruction and Rendering: Creating 3D point clouds from multi-view setups and rendering realistic 2D images across different viewpoints, leveraging among many approaches generative models, neural rendering and implicit models Human-in-the-Loop Approaches: Implementing active learning strategies that incorporate expert feedback at inference time, enabling real-time model correction and improvement with minimal labelling budget Field Evaluation: Conducting rigorous qualitative and quantitative evaluations of developed models on farm field experiments, integrating expert feedback to improve model performance Data Product Generation: Preparing comprehensive time series datasets of derived products, including raw data, 3D reconstructions, model estimations, and reference measurements for downstream analyses Software Development: Developing and maintaining codebases for the implemented methods, ensuring reproducibility, and facilitating future research and applications in the field of plant phenotyping Research and Development Design, develop and implement foundation model-based approaches for multi-trait plant phenotyping Extend and implement domain-specific and plant-specific, physiologically plausible, machine learning models Develop and evaluate domain transfer and adaptation methods for cross-platform phenotyping Design and deploy human-in-the-loop and active learning strategies Conduct field experiments and evaluate model performance in real-world field conditions Engage with diverse stakeholders including researchers, farmers, and breeders Collaboration and Scientific Communication Process and help curating large-scale multi-modal datasets from the FIP and field experiments Supervise and collaborate with students at different levels providing guidance and supervision Contribute to existing codebases and engage with open source communities Prepare publications for top-tier scientific venues Present research findings at conferences, seminars and workshops Communicate complex technical concepts to both expert and general audiences

The Swiss Data Science Center (SDSC) is a national research infrastructure in data science and artificial intelligence (AI) of the ETH domain, with EPFL and ETH Zurich as founding partners. Its mission is to support academic labs, hospitals, industry and public sector stakeholders, including cantonal and federal administrations, through their entire data science journey, from the collection and management of data to machine learning, AI, and industrialization. With a large multidisciplinary team of professionals across three locations (Lausanne, Zurich, Villigen), the SDSC provides expertise and services to various domains, such as health and biomedical sciences, energy and sustainability, climate and environment, and large-scale scientific infrastructures. The Swiss Data Science Center (SDSC) and the ETH Zurich’s Crop Science Group are seeking a Postdoctoral Researcher for the PhenoMix project, a Swiss National Science Foundation (SNSF) funded initiative.

• PhD in relevant field such as computer science, machine learning, data science, or domain science (e.g., plant phenotyping, agricultural sciences, environmental sciences) with demonstrated expertise in machine learning and computer vision

• Demonstrated research excellence through publications in relevant venues

• Strong background in machine learning and deep learning, particularly computer vision, with hands on experience in foundation models, transfer learning, domain adaptation

• Solid experience with modern deep learning frameworks (PyTorch preferred)

• Proven ability in scientific programming and prototyping in Python

• Ability to formulate research questions and design experiments independently

• Experience handling large and complex multi-modal datasets

• Excellent communication skills in English (written and oral)

• Positive attitude towards interdisciplinary collaboration

• Ability to work independently while contributing to team objectives

• Experience with 3D reconstruction techniques (structure from motion, neural rendering, etc.)

• Knowledge of active learning, human-in-the-loop, Bayesian optimisation

• Familiarity with agricultural sciences, plant phenotyping, or related domains

• Experience implementing, training and evaluating models for spatio-temporal data

• Interest in sustainable agriculture, crop science, or food safety challenges