During May 2026, Lilinus contributed to the dotnet/runtime repository by optimizing the TensorPrimitives.IndexOfMinMax functionality in C#. They developed a vectorized index-finding path using low-level programming and performance optimization techniques, introducing new interfaces to support cross-type and scalable data processing. Their approach included safe fallbacks for small input sizes to maintain correctness with minimal overhead. Lilinus addressed floating-point edge cases such as NaN propagation and signed-zero comparisons, improving reliability for analytics workloads. Through careful API design and refactoring, their work reduced future maintenance risk and enabled targeted performance profiling, establishing a robust foundation for safer and faster numeric primitives.