Literature shows that by controlling engines at extreme lean operating conditions (equivalence ratio <0.75) can reduce emissions by as much as 30% (Inoue, Matsushita, Nakanishi, & Okano (1993). Toyota lean combustion system—the third generation SAE, 930,873) and also it improves fuel efficiency by as much as 5-10%. However, the engine exhibits strong cyclic variation in heat release which may lead to instability and poor performance. A novel neural network (NN) controller is developed to control spark ignition (SI) engines at extreme lean conditions. The purpose of neuro-controller is to reduce the cyclic variation in heat release at lean engine operation even when the engine dynamics are unknown. The stability analysis of the closed-loop control system is given and the boundedness of all the signals is ensured. The adaptive NN does not require an offline learning phase and the weights can be initialized at zero or random. Results demonstrate that the cyclic variation is reduced significantly using the proposed controller developed using an experimentally validated engine model. The proposed approach can also be applied to a class of nonlinear systems that have a similar structure as that of the engine dynamics.