In the next decade, our understanding of exoplanet atmosphere will be revolutionized by the observations of 1000s of exoplanets thanks to a new generation of ultra-powerful telescopes: NASA-JWST, ESO-ETL, and ESA-Ariel. These novel observations are extremely complex and rich in information, pushing our current interpretation tools to their limits. The data reveals previously unseen physical processes that are difficult to jointly model, and require optimization over large free parameter spaces. Atmospheric retrievals, the most utilized inversion technique, are currently not equipped to handle the new scale of the datasets.
To meet these challenges, we propose to re-design one of the most utilized retrieval codes, TauREx. This project will incorporate recent community advancements, focusing on modernizing the computational and statistical framework to ensure performance and sustainability. Critically, technological advancements from the machine learning community will be deployed, establishing new standards for inversion problems in the field of exoplanet astronomy.
