Large networks composed of multiple input translinear elements (MITEs) have been typically limited by the mismatches between individual MITEs. This paper presents the methodology that allows for such systems to be feasible through the application of floating-gate programming techniques. We introduce designs for a programmable MITE, and demonstrate the ability to systematically reduce offsets through accurate programming of example circuits.