Accurate real-time upper level wind measurements can provide essential input into operational mesoscale models for their initialization and verification. Although there are a number of ground based wind profilers available (wind tracer lidar, doppler radar, and acoustical sounders), measuring upper level winds can be problematic and is highly dependent on favorable atmospheric conditions. In this study, we are investigating the use of a neural network to retrieve upper level winds from ground based wind profilers. Archived radiosonde measurements are used as the neural network training set. The idea is to retrieve timely and continuous upper level wind information from lower level wind observations that can be measured by commercial remote sensing wind profilers. To demonstrate the feasibility of this technique, a large training and testing set was extracted from the NCDC archived radiosonde data of North America (1946-1994) for a single location (El Paso, Texas, 32.0N 106.6W). The 700mb level winds were used as input into the neural network to derive the 400mb level winds in the preliminary test runs. Test results will be discussed and future directions will be recommended.