In this paper, we present our work on semantic deep mapping at scale by combining information from various sources and disciplines to study historical Amsterdam. We model our data according to semantic web standards and ground them in space and time such that we can investigate what happened at a particular time and place from a linguistics, socio-economic and urban historical perspective. In a small use case we test the spatio-temporal infrastructure for research on entertainment culture in Amsterdam around the turn of the 20th century. We explain the bottlenecks we encountered for integrating information from different disciplines and sources and how we resolved or worked around them. Finally, we present a set of recommendations and best practices for adapting semantic deep mapping to other settings.@en