Could you describe your role in the ODISSEE project?

The project aims to build a software package able to handle data processing workflows in the modern scientific landscape, which is characterised by overwhelming data volumes. One of the case studies is the SKAO. In this context, I am working to document and standardise our radio telescope data processing pipeline, which handles three data domains: sky distribution, antenna gains, and the measurements themselves.

My work will ensure that our pipeline can be used by other research infrastructures, by clarifying data formats and inputs/outputs, and by ensuring its interoperability with international standards – which often involves defining them! In this way, ODISSEE is already shaping, directly and measurably, the future of radio astronomy.

How is this pipeline organised?

A radio telescope measures sky brightness distributions (maps of the sky) indirectly. Furthermore, this indirect measurement is corrupted by a variety of effects, which can be astrophysical in nature (propagation effects such as Faraday rotation) or result from our instrumental hardware itself. These effects are measured along with the sky signal. Our processing aims to separate these contributions. In a radio astronomy data processing pipeline, there are thus three ‘boxes’ – SKY, MEASURE and ANTS – which separate the signal from the sky itself, its propagation effects, and our measurement space, in order to better correct the measurements and reconstruct the final image of the sky.

Astrophysical and instrumental propagation effects are modelled as antenna-based effects; we seek an estimate of the ANTS parameters from the measured data, then correct the observations to reconstruct SKY as best as we can.

What exactly does this clarification of formats you mentioned involve?

The initial DDFacet processing pipeline was developed by Cyril Tasse, an astronomer at the Paris Observatory, as part of his research work. Together with ODISSEE, he undertook a massive collaborative effort that allowed us to meet the data volume challenge. However, the pipelines remains a research work at present – to be adopted by other scientists and their infrastructures, there is a lot of groundwork to be done!

The first challenge is the lack of documentation and standardisation in the existing codebase, whose development emphasis was placed on the implementation and optimisation of new, cutting-edge features. As one of the first users of this software, I am coordinating the work to formalise and streamline the codebase. In practical terms, I analyse the existing code to understand exactly what each function does and when. Next, I define strict data formats that allow data and metadata to be stored as clearly as possible. Finally, I rewrite or adapt the code to use these formats, and I document each step so that users know exactly how to interact with the pipeline.

« Our current goals is to ensure that future user infrastructures using this code have a clear and well-documented interface. »

You started out as an astronomer and now you work with code. Is that more exciting?

I personally feel that my work is still fundamentally driven by astronomy itself : I want to fully understand the measurement and instrument to validate my understanding of the sky. In a sense, this is a bit like working as a line cook in the kitchen, to make sure that the dishes we serve are prepared to the highest standards. It only motivates me to eventually enjoy the meal !

My aim is to make this pipeline a recognised standard, used by research infrastructures around the world. If we succeed, researchers will be able to focus on scientific analysis without worrying about technical challenges.

Personally, I see this collaboration as an exceptionally effective dynamic, where everyone lifts each other up. I have rarely experienced such synergy between teams. Much of this effectiveness comes from Damien Gratadour, who knows how to motivate and challenge the teams, but it’s not just him: it’s also a team of competent, well-positioned and deeply motivated people.