1: \begin{abstract}

2: Current endeavours in exoplanet characterisation rely on atmospheric retrieval to quantify crucial physical properties of remote exoplanets from observations. However, the scalability and efficiency of said technique are under strain with increasing spectroscopic resolution and forward model complexity. The situation becomes more acute with the recent launch of the James Webb Space Telescope and other upcoming missions. Recent advances in Machine Learning provide optimisation-based Variational Inference as an alternative approach to perform approximate Bayesian Posterior Inference. In this investigation we developed a Normalising Flow-based neural network, combined with our newly developed differentiable forward model, \difftau{}, to perform Bayesian Inference in the context of atmospheric retrievals. Using examples from real and simulated spectroscopic data, we demonstrate the advantages of our proposed framework: 1) Training our neural network \comments{does not require a large pre-comupted training set and can be trained with} only a single observation; 2) It produces high-fidelity posterior distributions \comments{in excellent agreement with} sampling-based retrievals and; 3) It requires up to 75\%  \comments{fewer forward model calls} to converge to the same result. 4.) \comments{This approach allows formal Bayesian model selection.} \comments{We discuss the computational efficiencies of \difftau{} in relation to \taurex{}'s nominal forward model and provide a lesson's learned account of developing radiative transfer models in differentiable languages.} Our proposed framework contributes towards the latest development of a neural-powered atmospheric retrieval. Its flexibility and \comments{significant reduction in forward model calls required for convergence,} holds the potential to \comments{be an important addition to the retrieval tool box for large and complex data sets along with} sampling-based approaches.

3: \end{abstract}

4: