The characterization of radioactive waste in Swedish legacy waste drums is a critical task for nuclear safety and regulatory compliance. This project explores the feasibility of using machine learning to image actinide sources within waste drums by analyzing neutron and gamma emissions from spontaneous fission. High-performance computing (HPC) resources will be leveraged to generate large-scale Monte Carlo simulation datasets using Geant4, modeling the detection process with multiple detectors. The acquired simulation data will train a neural network to predict source locations based on cumulative time-difference distributions between gamma-ray and neutron detections. The proposed approach will be validated using experimental data to assess the model's accuracy and generalizability. By generating large datasets efficiently with HPC, this research aims to enhance non-destructive assay techniques for nuclear materials, contributing to improved radiation detection and safeguards.