IanMarteens/Austra

Austra is a technically substantial project: a .NET numeric library, a parser/formula language, a REPL, and a WPF application for experimenting with the language. The README makes the goal clear: provide optimized managed numeric algorithms for vectors, matrices, dates, DFT, time series, linear regression, AR/MA models, mean-variance optimization, simplex/linear programming, and related financial/statistical workflows. What makes it especially interesting is the performance angle. The library uses unsafe code and SIMD intrinsics, including AVX/AVX2/AVX512 paths, and the project files show packaging metadata, nullable enabled, documentation generation, trimming/AOT compatibility, and NuGet-oriented versioning for Austra.Library and Austra.Parser. Simulating usage, the best path is through the WPF app or REPL: type formulas, get evaluated results, and then bring the same capabilities into code through the library. This “library plus language” approach is powerful because it lets users explore before integrating. The improvement area is onboarding. The README explains the project well conceptually, but a new user would benefit from a minimal formula tutorial, a few copy-paste examples, package install instructions, and benchmark comparisons against Math.NET or native BLAS-like approaches. Given the advanced unsafe/SIMD code, visible tests and benchmark commands would also make the optimization claims easier to trust.