A two-zone NOx model intended for 1-D engine simulations was developed and used to model NOx emissions from a 2.5 L single-cylinder engine. The intent of the present work is to understand key aspects of a simple NOx model that are needed for predictive accuracy, including NOx formation and destruction phenomena in a DI Diesel combustion system. The presented two-zone model is fundamentally based on the heat release rate and thermodynamic incylinder data, and uses the Extended Zeldovich mechanism to model NO. Results show that the model responded very well to changes in speed, load, injection timing, and EGR level. It matched measured tail pipe NOx levels within 20%, using a single tuning setup. When the model was applied to varied injection rate shapes, it showed correct sensitivity to speed, load, injection timing, and EGR level, but the absolute level was well outside the target accuracy. The same limitation was seen when applying the Plee NOx model. Detailed CFD simulations showed that NO destruction is significant, occurs throughout the mixing controlled heat release, and can be up to 35% of the total formed NO. NO formation was confirmed to primarily occur in fuel lean regions around equivalence ratios of 0.95, and NO destruction was found to primarily occur in fuel rich regions stemming from the third Extended Zeldovich reaction. Fuel injection rate shaping changed the balance of NO formation and destruction in engine applications, but not in free-jet scenarios. Future paths focus on building the necessary fidelity into a computationally efficient NOx model, leveraging understanding from more detailed approaches.