Postdoctoral Researcher – Computational Electron Microscopy and Machine Learning

Descrizione

Testo originale dell'annuncio (in inglese)

Area of research: Scientific / postdoctoral posts Job description: Postdoctoral Researcher – Computational Electron Microscopy and Machine Learning The multidisciplinary Institute for Advanced Simulation - Materials Data Science and Informatics (IAS-9) brings together disciplines ranging from data analysis and machine learning to materials simulation, research data management and software development under one roof. In doing so, we extract new information from simulations and experiments, identify patterns and trends in microscopy data, and improve our understanding of why materials and processes work the way they do. We benefit from a strong connection to the Ernst Ruska-Centre for Microscopy and Spectroscopy with Electrons (ER-C) and to the Jülich Supercomputing Centre. Within IAS-9, the Deep Learning for Electron Microscopy group develops machine learning methods grounded in the physics of the measurement rather than in generic image statistics. We are looking for a postdoctoral researcher whose scientific home is electron microscopy itself and who wants to develop the computational methods the field now needs: identifying where microscopy questions genuinely require new methods, developing them, and validating them against real experimental data. Your Job This position sits at the interface between electron microscopy and machine learning. We are looking for a microscopist or physicist with substantial computational practice who wants to shape how machine learning is developed for electron microscopy – someone who can identify the open questions in the domain and carry them through to new methods. You will work in a team of data scientists and software developers, together with the doctoral researchers of the group, and in close collaboration with the Ernst Ruska-Centre (ER-C). Your tasks will include: Conducting research at the interface of electron microscopy and machine learning, including the application and adaptation of established methods, data analysis, and dissemination of results Identifying open questions in electron microscopy where progress is limited by analysis rather than by instrumentation, and where new machine learning methods therefore offer genuine scientific gain Developing, training and evaluating deep learning models for microscopy data – real-space imaging, 4D-STEM diffraction and in-situ time series – including the construction and characterisation of the experimental and simulated datasets your research requires Establishing what physically meaningful evaluation means for such models: what constitutes a correct answer, when an output is an artefact of the measurement, and which failure modes matter for the underlying materials science Conducting and interpreting simulation studies (e.g. multislice, Bloch-wave methods) that connect experiment, theory and model behaviour Participation in the scientific agenda of our collaboration with the ER-C: bringing microscopy questions into our method development, and turning model results into statements that are meaningful to microscopists Publishing in microscopy and materials science journals as well as at machine learning venues, and contributing to open datasets, benchmarks and software Co-supervising doctoral and master students, and contributing to proposal writing and to the group’s third-party funded projects Your Profile We are looking for a highly motivated colleague who is excited about new scientific endeavors with interdisciplinary approaches. For this you have: Completed Master's degree and PhD in Physics, Materials Science or a closely related field In-depth working knowledge of electron microscopy, including image formation and contrast mechanisms, electron diffraction, and substantial experience in at least one of the following: 4D-STEM, in-situ or operando microscopy, or quantitative HR(S)TEM. Experience with electron microscopy simulations (e.g. multislice or Bloch-wave methods), or with developing quantitative data analysis pipelines for