The use of machine learning methods to estimate causal effects—especially heterogeneous treatment effects—is currently an area of considerable study. We consider causal inference under the case where the number of treated units is small and the control group is large. For example, treated units may be those that underwent an experimental procedure and control units may be the set of units in a national database. For estimating treatment effects, Linus is developing mean-weighted case-specific random forests, in which bootstrap samples for random forests are weighted to heavily sample control units with similar characteristics to the treated units. Preliminary results show favorable performance compared to other methods.