Constitutive behavior models for steels are typically semi-empirical, however recently neural network is also being used. Existing neural network models are highly complex with a large network structure i.e. the number of neurons and layers. Furthermore, the network structure is different for different grades of steel. In the present study a simple neural network structure, 3:4:1, is developed which models flow behavior better than other models available in literature. Using this neural network structure constitutive behavior of 8 steels: 4 carbon steels, V and V-Ti microalloyed steels, an austenitic stainless steel and a high speed steel could be modeled with reasonable accuracy. The stress-strain behavior for the vanadium microalloyed steel was obtained from hot compression tests carried out at 850-1150 C and 0.1-60 s-1. It is found that a better estimate of the constants in the semi-empirical model developed for this steel could be obtained by simultaneous nonlinear regression. A model that can predict the effect of chemical composition on the constitutive behavior would be industrially useful for e.g., in optimizing rolling schedules for new grades of steel. In the present study, a neural network model, 5:6:1, is developed which predicts the flow behavior for a range of carbon steels. It is found that the effect of manganese is best accounted for by taking Ceq=C+Mn/6 as one of the inputs of the network. Predictions from this model show that the effect of carbon on flow stress is nonlinear. The hot strip mill at Jindal Vijaynagar Steel Ltd., Toranagallu, Karnataka, India, was simulated for calculating the rolling loads, finish rolling temperature (FRT) and microstructure evolution. DEFORM-2d a commercial finite element package was used to simulate deformation and heat transfer in the rolling mill. The simulation was carried out for 18 strips of 2-4 mm thickness with compositions in the range and 0.025-0.139 %C. The rolling loads and FRT could be calculated within 15 % and 15 C respectively. Analysis based on the variation in the roll diameter, roll gap and the effect of roll flattening and temperature of the roll showed that an error of 6 % is inherent in the prediction of loads. Simulation results indicated that strain induced transformation to ferrite occurred in the finishing mill. The microstructure after rolling was validated against experimental data for ferrite microstructure and mechanical properties. The mechanical properties of steels with predominantly ferrite microstructures depend on the prior austenite grain size, strain retained before transformation and cooling rate on the run-out table. A parametric study based on experimental data available in literature showed that a variation in cooling rate by a factor of two on the run-out table gives rise to only a 20 MPa variation in the mechanical properties.