In an effort to support the development of explicitly causal instead of correlative models, an open source library called CausalNex was launched that leveraged Bayesian networks to discover structural relationships in data. CausalNex has recently been extended to work for temporal data. The resulting Dynamic Bayesian Networks (DBNs) are powerful for predictive modeling because they model interactions on several timescales simultaneously: more immediate data dependencies can be captured within each time slice, and long-term dependencies can be captured between multiple time slices. To learn these networks, CausalNex uses an algorithm called DYNOTEARS, which formulates the structure learning as a continuous optimization problem subject to the constraint of acyclicity.