In this study, coffee ground (CG) was used to remove lead ions from aqueous solution. The physicochemical properties of CG determined were: point of zero charge (pHpcz 4.5), acid and basic group quantity (1.52 and 0.17meg/g, respectively), and specific surface area (<1m2/g). The pH effect (3–5) on lead adsorption was evaluated. An increase of solution pH causes an increment of adsorption capacity and the best adsorption capacity (q=22.9mg/g) was obtained at pH 5. The calculation of the Langmuir and Freundlich isotherm parameters was performed via stochastic optimization methods, such as Genetic Algorithm, Pattern Search and Simulated Annealing, and a gradient-based method. Pattern Search showed the fastest convergence and it generated feasible parameters for both isotherm models. On the other hand, Genetic Algorithm and Simulated Annealing, sometimes generated unfeasible parameters. The Langmuir and Freundlich isotherm models were compared against an Artificial Neural Network, which has the ability to learn unmodeled phenomena not considered by the previous models. Thus, an Artificial Neural Network has the best prediction capabilities, although it lacks the physical interpretation the isotherm models have.