My research sits at the intersection of econometrics, statistics, computer science, and numerical algebra, with a focus on models for time-series data. A large part of my recent work focuses on automatic model selection, which resulted in Autometrics. This machine learning method can handle more variables than observations, of particular interest in the form of `saturation' estimators. A specific example is impulse indicator saturation, where an impulse dummy is added for every observation. More information can be found in many papers, as well as the monograph written with David F Hendry: Empirical Model Discovery and Theory Evaluation. In recent years I have also started to research how these methods can help to create robust forecast methods. The most recent application is to short-term forecasting of COVID-19.