Donald Smith

Background

About Donald Smith

I build machine learning tools that get deployed. I solve user problems: First ask questions of the user or client to tease out the real issue underlying the client request. Keep that need front of mind. Go beyond a superficial understanding of the customer’s data. Analyze reality, not a simplified mental model. As a problem solver, I am not bound to one application area. I have worked lately with agentic LLMs. Before that, I applied deep learning to computer vision, time series to predictive maintenance, and stochastic models to network engineering. As a mathematician, I gravitate towards rigor and first principles. One component of rigor is articulating assumptions. First principles means knowing how machine learning models work. These attributes have made me very proficient at diagnosing flaws in system design, which, in turn, has led to numerous patents. I am a first-rate expositor, as evidenced by my best paper award and typically favorable referee reviews. I have an superb track record with teaching and mentoring. Most of my interns have returned for a second internship with me. 30+ publications and patents on a broad range of topics due to a talent for diagnosing and fixing the root cause of system defects.