Lithium-ion batteries (LIB) continue to permeate numerous sectors including electric vehicles, medical devices, and portable electronics due to their high energy densities. However, predicting the cycle life of LIBs remain challenging due to various factors including operational variability and fast charging requirements. Early cycle life prediction helps lower the cost of batteries through optimization of manufacturing processes, and thereby enhances cell life. In this context, machine learning techniques that synergistically combine physics-based data and experimental measurements hold the potential to detect underlying trends in capacity degradation. While most data-driven approaches require the utilization of high-rate tests to induce accelerated degradation, low currents pose challenges to cycle life prediction due to slower degradation onsets and longer feedback times. In this work, we develop a machine learning model to deconvolute the degradation response by feeding memory information from electrochemical signatures, enabling accurate prediction of cycle life and capacity loss in LIBs.