In this paper, we present a promising two-tier ensemble of heterogeneous machine learning models that integrates seven well-known machine learning classifiers to predict AMPs from macroalgae. The first tier of the ensemble consists of a suite of binary classifiers that identify AMPs from protein sequence data which are then forwarded to a second tier multi-class ensemble to characterise their functional family type. The two-tier ensemble was successfully used to identify 39 putative AMP sequences in twelve macroalgae species from three different phyla groups. The approach we describe is not limited to AMPs and can also be applied to search sequence data for other types of proteins.