Cost, performance, and reliability are requirements of many embedded software programs. Both sophisticated and simple applications can be developed quickly without resorting to a Real-Time Operating System (RTOS). This paper discusses reasons why an RTOS may not be an appropriate choice. In this paper, an application employing a high-speed neural network and a machine-learning algorithm is developed as a framework to describe foreground/background tasking, re-entrant code, and interrupt sequencing.