mfagerlund 

A minimal, dependency-free fluent builder for generating SVG documents (shapes, paths, text, clip paths and SMIL animations) in .NET. Uses System.Numerics.Vector2; no third-party dependencies.

Nano-sized nonlinear least squares (Levenberg-Marquardt) for .NET — dense and sparse, analytic or finite-difference Jacobians, plus an L-BFGS minimizer. No autodiff, one small dependency (CSparse).

A Godot-free, batched-PPO reinforcement-learning core for .NET, built on the Tensotron tensor + autograd engine. Multi-channel ControlSpec, in-process training and inference, byte-exact model save/load — no... More information

A PyTorch-faithful, float32 tensor and autograd library for .NET, GPU-accelerated with ILGPU. Correctness-first: if it doesn't behave like PyTorch, it's a bug.