A numerical method for designing chemical vapor deposition (CVD) reactors is proposed, which consists of three steps: extraction of the chemical mechanism and reaction rates using small-scale, well-defined experiments; prototyping a large-scale reactor using this chemistry combined with computational fluid dynamic (CFD) simulations (we call these virtual experiments); and experimental optimization of a prototype reactor designed from the virtual experiments. This design methodology was validated using the model CVD process of low-pressure CVD (LPCVD) of tungsten silicide (WSix) from WF6 and SiH4 in the temperature range of 130-360 degrees C. A tubular hot-wall reactor was used as the small-scale reactor for diagnosing the gas-phase and surface chemical mechanism. The CFD code FLUENT was used for the numerical simulations, and no fitting parameters were used in chemical reaction mechanisms. Simulations of the radial WSix film growth rate and Si/W composition profiles on a 5 in. wafer in a cold-wall reactor agreed well with experimental measurements. Such comparisons indicate that numerical reactor design can replace currently used empirical design methods.