A variational algorithm has been developed to retrieve atmospheric humidity profiles from satellite observed radiances from the microwave sounder SAPHIR aboard Megha-Tropiques. In order to speed up the retrieval process, an ANN based fast forward model for simulating the upwelling radiances, corresponding to 6 channels in SAPHIR has been constructed this was found to be very efficient and accurate. The minimization of the cost function in 1-Dimensional Variaional framework was accomplished using the Levenberg-Marquardt algorithm. The retrieved profiles were compared with SAPHIR Level 2 data obtained from MOSDAC, Indian Meteorological Department, India. A good agreement was observed between the two. © 2015 IEEE.