Current research efforts in combustion science and technology largely focus on the development of techniques that can model reacting flows in complex geometries accurately.  The use of computational fluid dynamics and properly reduced kinetic models is a viable solution for the present and the foreseeable future.  Such efforts, however, can be of limited value if the associated models are incomplete and/or inaccurate.  Chemical kinetic models that describe the fuel pyrolysis and oxidation can be rather large, and the rate constants are often associated with large uncertainties.  The nonlinear nature of the problem and the large dimensionality associated with the underlying inverse problem make it a exceedingly difficult to arrive at a reliable kinetic model for combustor simulations.  The talk will focus on recent advances in the computational techniques developed for developing predictive combustion chemistry models. The focus will be on a systematic, hierarchical approach to reduce the dimensionality and uncertainties in the model.  Several topics will be discussed, including (1) application of ab initio electronic structure theories to explore the potential energy surfaces of intramolecular interactions and its application in the thermochemistry of combustion intermediates; (2) application of reaction rate theories for the rate parameters of elementary reactions critical to combustion simulations; (3) detailed kinetic modeling for the combustion of jet fuel surrogates; and (4) systematic model optimization, uncertainty quantification and propagation.