The breakthrough of data mining in the mid-nineties can be explained as the culmination of several independent developments: the convergence of machine learning research, the maturity of data base technology, the workstation and client server-revolutions, the rapid growth of decentralized data collection, the stabilization of administrative systems in large organizations, the apparent failure of traditional marketing techniques. I will look back at these developments from my perspective as R&D manager/CEO at Syllogic and later Perot Systems. We started with ambitious AI projects in data base environments (OBIS, CAPTAINS) but we soon expanded our interest to the analysis of dynamic systems (ASM, ICT, JSF, Robosail). In the new Millennium, research turned its attention to e-Science, (VL-e) large RDF data bases and Big Data (Commit). My current research interest lies with complexity measures for large data sets and methodology for e-Science in the 21st century (Atlas of Complexity). Using examples from my R&D practice, I will give an overview of this period of nearly 30 years of research and sketch some perspectives for the future.