Distributed computing based on the Master-Worker and PULL interaction model is applicable to a number of applications in high energy physics, medical physics and bioinformatics. We demonstrate a realistic medical physics use-case of a dosimetric system for brachytherapy using distributed GRID resources. We present the efficient techniques for running parallel jobs in a case of the BLAST, a gene sequencing application, as well as for the Monte Carlo simulation based on Geant 4. We present a strategy for improving the runtime performance and robustness of the jobs as well as for the minimization of the development time needed to migrate the applications to a distributed environment. © 2004 IEEE.