Why I do it?
The natural and physical sciences are on the brink of transformation; mainly because the ability to collect data systematically has exploded. The intuitive and (often) deterministic models of systems are being replaced by abstract models of 'data'. In physics for instance, the discovery of exotic particles is largely dependent on the statistical prowess of scientists working on these experiments rather than veteran physics ability. Machine learning offers practical algorithms to parse large volumes of data with the aim of encoding their hidden characteristics. The classical models of cause and effect are deeply limited in their ability to encode intrinsic relationships within the data and account for their evolution. ML is relevant in any domain that is data rich. Amongst its most successful applications include - discovery of the Higgs, gene sequencing, classifying astronomical objects, medical diagnosis. I enjoy it because it means I can work with scientists in different domains and get to ask them fundamental questions about their data and science.
The 'cult of AI' is not only transforming science. In financial markets, predicting the future is the holy grail. It is no secret that if there is anything that comes close to that aim, it is machine learning. There is a severe knowledge gap in quantitative finance and a dearth of people who can straddle the two spheres.
Besides the extensive utilitarian benefits of training machines to do a wide variety of useful things in science the need for people who can unravel and explain the workings of learning algorithms is ever on the rise.