A robust optimization scheme, known as rkmGen, for reaction rate parameter estimation has been developed for the generation of reduced kinetics models of practical interest for reactive flow simulations. It employs a stochastic optimization algorithm known as simulated annealing (SA), and is implemented in C++ and coupled with Cantera, a chemical kinetics software package, to automate the reduced kinetic mechanism generation process. Reaction rate parameters in reduced order models can be estimated by optimizing against target data generated from a detailed model or by experiment. Target data may be of several different kinds: ignition delay time, blow-out time, laminar flame speed, species time-history profiles, and species reactivity profiles. The software allows for simultaneous optimization against multiple target data sets over a wide range of temperatures, pressures, and equivalence ratios. In this paper, a detailed description of the optimization strategy used for the reaction parameter estimation is provided. To illustrate the performance of the software for reduced kinetic mechanism development, a number of test cases for various fuels were used: one-step, three-step, and four-step global reduced kinetic models for ethylene, Jet-A and methane, respectively, and a 50 step semiglobal reduced kinetic model for methane. The 50 step semiglobal reduced kinetic model was implemented in the Star*CCM+ commercial CFD code to simulate Sandia Flame D using laminar flamelet libraries and compared with the experimental data. Simulations were also performed with the GRI3.0 mechanism for comparisons.