mesh.applied

A memory-efficient, single-GPU training and inference architecture for formalizing natural language mathematics into Lean 4 statement code using quantized Mixture-of-Experts, persistent Lean REPL compiler feedback, and dual-tier retrieval augmentation.

Variational quantum circuits inside PINNs for 2D Navier–Stokes against matched classical baselines. The shipped result is a classical HarmMLP that hits 1.29% FD-curl ω error on a Taylor-Green vortex (1.78% on a ~2.5× faster distilled net).