Exoplanet Atmosphere Characterization

This project intends to develop cutting-edge machine-learning tools for spectral analysis to characterize the atmospheres of exoplanets. The project will explore various state-of-the-art and recent techniques to build a robust and general model with strong regressional properties. The resultant models from our study will be benchmarked against pre-existing baselines based on the principles of statistics and atmospheric sciences. Additionally, relevant tutorials and documentation will be provided, keeping future developments in mind. We propose simulation-based inference, causal representation learning and domain-adapted transformers for atmospheric retrievals.