For high-accuracy modeling of light propagation in biological tissue, the Monte Carlo method is currently the best tool in our arsenal. Its principal drawbacks include relatively slow convergence and, in the case of 3D problems, prohibitive computer memory requirements. Furthermore, in most problems, the number of iterations required to obtain satisfactory convergence varies substantially with distance from the light source; obtaining useful information at large distances from the light source requires that we generate more data than needed at smaller distances. We will present our work on an adaptive technique, which trades off spatial resolution in regions of low light intensity against the two aforementioned drawbacks to generate an `optimal' intensity map from the available data. With this technique, memory requirements scale with the degree of detail required, and not with the physical size or dimensionality of the problem. The reduced memory requirements make 3D problems tractable, and because the technique is adaptive, a generic approach is applicable to virtually any problem, without need to tailor the program to particular geometries.