LLM training in simple, raw C/CUDA

Tile primitives for speedy kernels

Templated C++/CUDA implementation of Model Predictive Path Integral Control (MPPI)

3D Gaussian Splatting in JAX

Code supporting the WAFR paper "A Performance Analysis of Differential Dynamic Programming on a GPU," and the ICRA workshop follow on work deploying the algorithm onto robot hardware.

nvblox Torch

Fork of the ops from Erik Wijman's Pointnet++ repo

Kinodyanmic Parallel Accelerated eXpansion