There are several machine learning skills that are in high demand in the global marketplace today.
The skill most required is the ability to come up with fundamental innovations in machine learning, and implement them to solve practical problems. For a research career in AI, you need a PhD, preferably from a well-known programme, and research competence as demonstrated by published papers, implemented solutions and peer acceptance. For those at the forefront of research, the sky is the limit, and seven-figure USD salaries are not infrequent.
The next tier of demand is for people who can build practical implementations, especially in collaboration with a cutting-edge research team. For a career as a machine learning implementation expert, you need substantial work experience and strong software development skills. You need to be able to build production quality systems for industrial big data, and not just know a bit of R or Python that lets you put together a model on a toy dataset. However, a lot of effort today is directed at training “machine learning engineers” who have this superficial level of expertise.
Ever since big data and machine learning became highly hyped buzzwords some five years ago, many STEM graduates have attempted to enter the field of machine learning by retraining themselves through short courses, MOOCs, or coding boot camps. This channel of career advancement is nearing saturation levels, and opportunities for such repurposed machine learning engineers are beginning to dry up.