The silicon and scintillator sensor-based high-granularity calorimeter (HGCAL) is a part of phase 2 upgrade to the existing Compact Muon Solenoid end-cap calorimeter at Large Hadron Collider. The HGCAL data will be used to obtain the direction and measure the energy of electrons, positrons, photons, hadrons, and jets in the end-cap region. Electron, while losing its energy in the calorimeter, creates a shower of secondary particles spreading out in all directions in the detector material. The energy loss and lateral and longitudinal pattern of these particles within the HGCAL provide valuable information to extract the initial direction and energy of electrons. Graph neural networks (GNNs) is a subclass of algorithms belonging to the geometric deep learning (GDL) that facilitates understanding data with inherent geometric structures, which is not possible with traditional neural networks. GNNs excel in processing complex data as they directly model relationships between data points represented as nodes in a graph. This chapter presents the novel approach of predicting the electron energy produced at the collision point (vertex) and passing through the tracker followed by HGCAL in the end-cap region using GNN. The energy is predicted using a regression model, and the results demonstrated the effectiveness of GNN in accurate prediction of the electron energy. Hits recorded in the HGCAL detector for electron events simulated in the Pt range between 25–250 GeV at collision point have been used for training and testing. The test sample was reconstructed using these hits with an overall energy resolution of 2%.