Omar Khan

We present UMI (Unified Median Iterative), a novel robust location estimator for detrending photometric time series in exoplanet transit surveys. UMI modifies the standard Tukey bisquare M-estimator with two innovations: (1) an asymmetric weight function that penalizes downward deviations (transit dips) more aggressively than upward ones, exploiting the physical constraint that transits are always below the stellar continuum, and (2) an upper-RMS scale estimator computed from above-median residuals only, ensuring that transit dips never contaminate the noise estimate. Implemented as a fused HIP/CUDA GPU kernel, UMI achieves 69x faster detrending (3.4 ms vs 234 ms per star) and 37x faster full pipeline throughput compared to the standard wotan's biweight. On 1000 real TESS stars, UMI reduces the median per-star depth recovery error at 0.1% transit depth from 20.5 (biweight) to 15.8%, a 23% improvement. On Kepler, where lower photometric noise allows the asymmetry to work more effectively, the improvement grows to 71% (4.2% vs 14.6%). Validated on 802 confirmed exoplanets across both missions, UMI recovers more planets than biweight, Welsch, and Savitzky-Golay combined. The tool is publicly available as pip install torchflat.