Young Graduate Trainee for Generative Modelling for Space Data

• The Advanced Concepts & Studies Office ensures the overall coordination, coherence and performance of programme and corporate studies in support of the preparation of the Agency’s future activities in line with its long-term strategic objectives and priorities. This includes the Advanced Concepts Team (ACT), ESA’s in-house research think-tank.

  The ACT performs beyond-the-horizon multidisciplinary research for space, exploring new approaches to space-related R&D (including competitions, prizes, games), research for disruptive innovation, developing an expert network at academic level and providing a capability for fast first-look analysis of problems, challenges and opportunities. The ACT engages in collaborative research relations with university institutes and research centres, focusing on advanced research topics of potential strategic interest to the space sector and experimenting with new forms of teamwork. To achieve its goal, a multidisciplinary research environment is provided in which young scientific and engineering post-doctoral and postgraduate researchers work on emerging technologies and innovative concepts.

  More information on the Team is accessible on the ACT website (www.esa.int/act ). Candidates interested are encouraged to visit the ESA website: www.esa.int

You will carry out most of the activity in the field of generative modelling, in particular exploring potential synergies between generative modelling and space relevant data sets with a particular emphasis on, but not limited to, guidance control & navigation. Generative modeling is intended here as an unsupervised learning task able to automatically discover and learn patterns in input data: the resulting generative model can then be used to output new examples that plausibly could have been drawn from the original data set. One obvious use in a space context is to learn how to generate real data from synthetic data, filling the “reality gap” that is particularly problematic in space science applications. In the case of a pose estimation problem (here just as an example), this would entail generating realistic textures and illumination conditions out of a synthetically generated (on ground) image of orbiting spacecraft.

Tasks will depend on your expertise and motivation and may include:

• applications of Generative Adversarial Networks to pose-estimation.
• using world-models to generate new scenarios/examples which are "imagined" by the prediction model, in the context of machine vision for autonomous robots for instance.
• using generative models to augment Kelvins datasets.
• organising a Kelvins competition on generative models.
• proposing a new space-related application for a machine-learned generative model.
• using generative models for creating plausible physical systems. e.g. magnetohydrodynamics of Earth-Sun interaction, estimating planetary parameters of a star system based on incomplete data
• generative models in the context of text and information retrieval.
• using generative models for domain adaptation.

 
Depending on the nature of the project, you may be required to interface with the academic community in these fields. You will be a full member of the ACT and therefore expected to contribute to the development and assessment of new concepts and technologies for space applications in close interaction with ACT researchers who work on a broad range of disciplines, including: informatics, artificial intelligence, climate modelling, energy systems, fundamental physics, biomimetics, computational management science and mission analysis. Based on your detailed background and interests and the Team opportunities and needs, you will be involved in a number of other ACT initiatives including studies conducted via the Ariadna scheme (www.esa.int/ariadna ) and will also participate in reporting and communicating Team results internally and externally.

You should have just completed, or be in the final year of a university course at Master's level (or equivalent) in a technical or scientific discipline in particular in artificial intelligence, computer science, informatics or related fields.

You should have a good background and strong interest for machine learning including deep learning and evolutionary algorithms including genetic programming and neuroevolution. Excellent programming experience (e.g. C++, C, Python) is required. Experience with GPGPU and web-technologies is an asset.

The working languages of the Agency are English and French. A good knowledge of one of these is required. Knowledge of another Member State language would be an asset.　

You should demonstrate good interpersonal skills and the capacity to work both independently and as part of a team. 　

During the interview your motivation and overall professional perspective/career goals will also be explored.