A number of upcoming surveys in different wavebands are expected to probe, with unprecedented sensitivity, the epoch of cosmic reionization when the first stars in the Universe formed. Interpreting the large and complex data sets from these observations would require theoretical models that are extremely detailed and, at the same time, computationally efficient. It turns out that, although tremendous progress has been achieved towards building such models, they are still not at the level as one would require to interpret the extremely high-quality data expected from the upcoming probes. To address this shortcoming, we are in the process of developing computationally efficient physical models of the high-redshift universe. This talk will review some interesting aspects of these semi-numerical models keeping in mind various ongoing and upcoming observational efforts.