Kavach-Opt | Indigenous GPU-Accelerated Optimization Solver | SIH 2026 | Team S8ul

Kavach-Opt is a from-scratch C++20 mathematical optimization solver developed by Team S8ul for Smart India Hackathon 2026. The solver supports Linear Programming (LP), Mixed-Integer Linear Programming (MILP), Convex Quadratic Programming (QP), and Convex Mixed-Integer Quadratic Programming (MIQP) through a shared sparse numerical core. This demo showcases the actual compiled C++ solver, the connected Kavach-Opt workstation, independent mathematical verification, benchmark validation, CPU multicore execution, scoped CUDA support, and an illustrative refinery optimization case study. Key components demonstrated: From-scratch C++20 optimization engine Sparse CSR/CSC numerical architecture Two-Phase Revised Simplex and Dual Simplex Branch-and-Cut for MILP Active-Set method for Convex QP Independent post-solve verification CPU Sequential and Multicore execution Seven custom CUDA kernels with automatic CPU fallback 13 Netlib LP instances independently verified 23-instance cross-validation against HiGHS 412 automated tests Illustrative refinery crude blending and capacity optimization Important: The refinery dataset shown in this demonstration is illustrative/synthetic and is not MRPL operational data. CUDA kernels are implemented, but GPU speed benchmarking on NVIDIA hardware has not yet been performed. Team: Team S8ul Team ID: 172336 Problem Statement: SIH26119 - Indigenous GPU-Accelerated Optimization Solver