This research aims to optimize deep learning models constituting long short-term memory and dense neural networks using the genetic algorithm (GA). This novel scenario has been applied to automatically identify reservoir types (homogeneous and natural fracture) and their external boundaries (infinite acting, circularly closed, and constant pressure) and estimate the related parameters. The suggested scenario includes two classifiers and 48 predictors to handle reservoir/boundary model detection and parameter estimation simultaneously. This methodology can recognize the reservoir/boundary models and predict wellbore storage constant, storativity ratio, skin factor (S), CD (dimensionless wellbore storage constant) × e2S, and inter-porosity flow coefficient. The pressure signals required for training the classifier and predictor models have been simulated by solving governing equations with added noise percentages. The hyperparameters of the intelligent models have been carefully tuned using the genetic algorithm to improve their classification/prediction accuracy. The GA-optimized classifier attained 94.79% and 94.29% accuracy over the training and testing groups of the pressure transient signal, respectively. The separately trained 24 optimized predictors converged well to estimate the reservoir parameters. The reliability of the proposed scenario has also been validated using two actual-field well-testing signals. The results indicate that the suggested procedure accurately identifies the reservoir/boundary model and efficiently approximates the associated parameters.