A user is seeking a reliable curve-fitting routine in C or C++ and currently uses the Levenberg-Marquardt method, which has been failing frequently. Suggestions indicate that adjusting fitting parameters in this routine could enhance its robustness. The idea of employing a genetic algorithm to optimize these parameters is proposed, but the user lacks expertise in genetic algorithms. They are looking for recommendations on accessible code or packages to implement genetic algorithms effectively. The discussion emphasizes the need for improved fitting techniques for challenging data sets.