The design principles and main characteristics of the MWP967KV ground based 35 channel microwave profiling radiometer are presented. The prototype radiometer is a compact integration of dual band (K band and V band) broadband radiometry antennas, two receivers, a synthesizer local oscillator, a surface meteorological sensor, and a power supply in a cabinet. It has the capability of detecting downwelling atmospheric thermal emission continuously on a minute time scale. Several levels of observed data are composed and regulated. By using BP artificial neural network retrieval algorithms, the radiometer is capable of retrieving the vertical distributions (profiles) of atmospheric temperature, water vapor, relative humidity, and other variables in real time. The retrieved profiles range from the surface up to 10 km. The radiometer was used in atmospheric observation in Beijing in autumn to winter of 2012 for about two months. The data of 69 radiosonde cases are gathered to evaluate the radiometer’s performance. The accuracies of dual band atmospheric bright temperature values are analyzed. The retrieved temperature and humidity profiles are analyzed as well, and the mean difference, root mean square difference, and correlation are calculated. The statistical results indicate that the radiometer has the ability to support real time weather surveillance.