Mega-pixel charge-integrating detectors are common in near-IR imaging applications. Optimal signal-to-noise ratio estimates of the photocurrents, which are particularly important in the low-signal regime, are produced by fitting linear models to sequential reads of the charge on the detector. Algorithms that solve this problem have a long history, but can be computationally intensive. Furthermore, the cosmic ray background is appreciable for these detectors in Earth orbit, particularly above the Earth's magnetic poles and the South Atlantic Anomaly, and on-board reduction routines must be capable of flagging affected pixels. In this paper we present an algorithm that generates optimal photocurrent estimates and flags random transient charge generation from cosmic rays, and is specifically designed to fit on a computationally restricted platform. We take as a case study the Spectro-Photometer for the History of the Universe, Epoch of Reionization, and Ices Explorer (SPHEREx), a NASA Small Explorer astrophysics experiment concept, and show that the algorithm can easily fit in the resource-constrained environment of such a restricted platform. Detailed simulations of the input astrophysical signals and detector array performance are used to characterize the fitting routines in the presence of complex noise properties and charge transients. We use both Hubble Space Telescope Wide Field Camera-3 and Wide-field Infrared Survey Explorer to develop an empirical understanding of the susceptibility of near-IR detectors in low earth orbit and build a model for realistic cosmic ray energy spectra and rates. We show that our algorithm generates an unbiased estimate of the true photocurrent that is identical to that from a standard line fitting package, and characterize the rate, energy, and timing of both detected and undetected transient events.