We have developed a new computational framework for remote-sensing data analysis. This framework has two main elements: 1) Rigorous forward physical modeling of the interaction of electromagnetic radiation with an absorbing and scattering (2D and/or 3D) medium. 2) Bayesian estimation, from sensor data, of parametric representations of medium properties, as well as those of the source and the sensor. In its present form, the framework uses discrete-ordinates radiative-transfer theory to model the propagation of solar radiation in the atmosphere and its reflection from the surface in the plane-parallel geometry. Molecular and particulate interactions are modeled with the help of HITRAN database and Mie-scattering code; surface reflectance is represented by one of several standard models suitable for the geo-location. Sensor output is simulated using appropriate instrument field of view and spectral response functions. The code also produces forward-model derivatives with respect to unknown parameters of interest (vertical profiles of temperature, gas concentrations, and cloud and aerosol microphysical properties; spectral and bi-directional surface reflectance; etc.), which are then used by a Gauss-Newton/Levenberg-Marquardt routine for retrieval from data. Work is under way to build atmospheric dynamics into the framework so that sequences of hyper-spectral images can be assimilated via spatio-temporal Kalman filtering. The framework can be readily extended for use with active sources (radar, lidar), as well as passive sensing in other parts of the spectrum (infrared, microwave). We are interested in applying this tool to the characterization of the atmospheric and surface composition and dynamics of Earth, Mars, outer planets in our system, and distant planets around other stars. It may also be used in the design of new instruments, both as a tool for sensitivity analysis and as a way of enabling autonomous (i.e., self-monitoring and calibrating) sensor operation. TRL: 4-5 (tested on multi- and hyper-spectral satellite data sets)