WEHI’s imaging facilities have built a 2–3 petabyte microscopy archive, including 3D data from confocal, light-sheet and lattice light-sheet microscopy. Foundation models are transforming image analysis, but progress has been largely 2D, as large 3D collections are rarely available for training. Making it AI-ready is an ongoing effort, with several ways to contribute:
- Curating datasets, pretraining or benchmarking foundation models
- Exploring whether models trained across the archive generalise across instruments and modalities, including generative models to augment scarce annotations
- Profiling the archive and extending our metadata framework (REMBI, RO-Crate), including LLM-assisted extraction from proprietary formats
- Reusing archived data to answer new biological questions
Students will develop skills in large-scale data engineering, machine learning and scientific data standards. Python experience expected, microscopy experience not required.
This project is open to students in the Medical Student Research Internship program.
