A new golden age in astronomy is upon us, dominated by data. Large astronomical surveys are broadcasting unprecedented rates of information, demanding machine learning as a critical component in modern scientific pipelines to handle the deluge of data. The upcoming Legacy Survey of Space and Time (LSST) of the Vera C. Rubin Observatory will raise the big-data bar for time- domain astronomy, with an expected 10 million alerts per-night, and generating many petabytes of data over the lifetime of the survey. Fast and efficient classification algorithms that can operate in real-time, yet robustly and accurately, are needed for time-critical events where additional resources can be sought for follow-up analyses. In order to handle such data, state-of-the-art deep learning architectures coupled with tools that leverage modern hardware accelerators are essential.

The work contained in this thesis seeks to address the big-data challenges of LSST by proposing novel efficient deep learning architectures for multivariate time-series classification that can provide state-of-the-art classification of astronomical transients at a fraction of the computational costs of other deep learning approaches. This thesis introduces the depthwise-separable convolution and the notion of convolutional embeddings to the task of time-series classification for gains in classification performance that are achieved with far fewer model parameters than similar methods. It also introduces the attention mechanism to time-series classification that improves performance even further still, with significant improvement in computational efficiency, as well as further reduction in model size. Finally, this thesis pioneers the use of modern model compression techniques to the field of photometric classification for efficient deep learning deployment. These insights informed the final architecture which was deployed in a live production machine learning system, demonstrating the capability to operate efficiently and robustly in real-time, at LSST scale and beyond, ready for the new era of data intensive astronomy.