This is a simple parallel quadratic program solver in Jax (with full history search, although it is not memory-optimized). It implements a Newton descent algorithm with backtracking line search.
Project done as part of the course Convex Optimization by Alexandre d'Aspremont.