We integrate high-precision aftershock locations with geodetic inverse modeling to create a more complete kinematic model for the Kozani-Grevena earthquake sequence. Using the double-difference algorithm, we have improved relative hypocentral locations by a factor of ∼7 and thus imaged the details of the fault network associated with the seismic sequence. The interpreted fault network consists of multiple segments including (1) a master normal fault that strikes nearly due west and dips toward the north at 43°, extending from 6 to 15 km depth; (2) an upper segment that connects the top of the seismicity to the observed surface ruptures and dips 70°; (3) hanging wall antithetic faults; (4) a more steeply dipping southwest striking linking structure at the southwest end of the rupture; and (5) a separate south dipping segment at the southwestern end of the aftershock cluster. The imaged fault segment dimensions, orientations, and geometric relationships are consistent with regional fault patterns. Using slip inversion on triangular dislocation patches, we calculate variable slip on the imaged three-dimensional fault network that best fits the surface displacements observed by satellite interferometric synthetic aperture radar (InSAR). In our preferred model we find that the majority of slip occurred at depth on the west and southwest striking segments. By comparing these results to a planar fault model derived solely from the InSAR data using nonlinear inversion methods we demonstrate that the three-dimensional model improves the fit to the geodetic data while incorporating the observations of surface rupturing and aftershock distributions.