Clostridium novyi causes necrotic hepatitis in sheep and cattle, as well as gas gangrene. The microorganism is strictly anaerobic, fastidious, and difficult to cultivate in industrial scale. C. novyi type B produces alpha and beta toxins, with the alpha toxin being linked to the presence of specific bacteriophages. The main strategy to combat diseases caused by C. novyi is vaccination, employing vaccines produced with toxoids or with toxoids and bacterins. In order to identify culture medium components and concentrations that maximized cell density and alpha toxin production, a neuro-fuzzy algorithm was applied to predict the yields of the fermentation process for production of C. novyi type B, within a global search procedure using the simulated annealing technique. Maximizing cell density and toxin production is a multi-objective optimization problem and could be treated by a Pareto approach. Nevertheless, the approach chosen here was a step-by-step one. The optimum values obtained with this approach were validated in laboratory scale, and the results were used to reload the data matrix for re-parameterization of the neuro-fuzzy model, which was implemented for a final optimization step with regards to the alpha toxin productivity. With this methodology, a threefold increase of alpha toxin could be achieved.