Metastability characterizes the long transient relaxation towards a stable state of certain dissipative dynamical systems, including quantum ones. Interestingly, metastable behaviour has been described in the interaction between neurons in the brain and it has been shown as a mechanism to have associative memories. The aim of this project is to explore the effect of this phenomenon in quantum reservoir computing, a machine learning approach inspired by neuromorphic computation. We will focus on a nonlinear oscillator with dissipation in the formalism of quantum master equations, we will design a suitable reservoir computing architecture, and finally we will address the performance of this machine learning approach for time series processing in the metastable regime.