Worked on the astropy/astropy repository to address numerical stability in statistical utilities, focusing on improving the reliability of Poisson confidence interval calculations for large sample sizes. Tackled a bug where the existing method could overflow by refactoring the Kraft-Burrows-Nousek approach to use the gammaln function, ensuring stable results in high-N regimes. Added a regression test to validate the fix and prevent future regressions, enhancing the robustness of the statistics module. Utilized Python and applied skills in numerical analysis and scientific computing to deliver a targeted solution that improves accuracy for users relying on statistical computations within Astropy.