The decarbonization of internal combustion engines (ICE) necessitates advanced predictive tools for ammonia, a zero-carbon fuel with complex combustion dynamics. This study develops a deep neural network (DNN) framework to predict performance parameters and emissions in ammonia-fueled spark ignition engines. Experimental data from a V-twin engine and AVL-BOOST simulations were integrated to train a multi-layer DNN architecture. Data partitioning followed a 70%-15%-15% split for training, validation, and testing. The model achieved exceptional accuracy, with training/validation losses converging range from 10–4 to 10-3, MAE below 0.5%, and R2 > 0.99 across all datasets. The DNN captured critical non-linear phenomena: BMEP’s dependence on ignition timing, showing peak performance at optimal phasing, and NOx reduction under advanced ignition due to lowered peak temperatures.