LOKI-G is a machine learning project designed for physical machines, enabling edge AI capabilities. It adapts the LOKI algorithm for training via imitation and reinforcement learning using Closed-form Continuous-Time Neural Networks and Neural Circuit Policies.