Research Fellow

This candidate will have a background in image data compression, supported by experience with AI and machine learning. They will utilise this knowledge to develop domain specific compression by exploiting spatial and temporal redundancies, including through implementation of AI. They will also create the encoder/decoder software for compressed data, so require experience of coding these structures in C/C++.

X-ray Computed Tomography enables non-destructive observation of an objects internal structure. From material phases of novel alloys to uncovering a fossils provenance it supports a broad spectrum of academic research. However, even modest operation of a single system generates >10TB of raw data annually which requires archiving for 10+ years. This had led to an unsustainable data accumulation globally, desperate for a solution.

This project will create the first accessible, domain specific, machine independent CT data compression methods. Your experience with image data compression structures combined with intuition of domain specific redundancies will enable us to develop such methods, slashing storage requirements by up to 80% and directly lowering the carbon footprint of global research infrastructure. You will initially focus on lossless compression with predictor models and inspiration from video encoding. Stronger compression ratios will be identified through machine learning and AI methodologies. Subsequently, AI supported lossy methods will be developed with a focus on optimal compression with minimal acceptable data loss. These methods will be integrated into user friendly open-source compression software so everyone can benefit.

The project is joint with University of Cambridge and University of Portsmouth. It includes an annual two month secondment to Cambridge with all travel covered, to receive the experience of multiple research environments to develop the best data compression.

There are a series of project partners including NASA, Rolls-Royce, and JLR who will be appraising the research as it evolves, ensuring the outcomes are exploited.

An experienced data compression engineer with experience in AI and machine learning. A keen interest in identifying patterns in data, and learning from fields adjacent to their knowledge.

This position requires an understanding of existing image and video codecs, with an intuition of how they would be adapted specifically for these datasets. AI methodologies for compression will be a key focus to identify further refinements. Previous experience in creating software and management thereof is required.

The candidate must have Python and C/C++ programming experience. No previous experience with X-ray Imaging is necessary.