SIH26119: Indigenous GPU-Accelerated Optimization Solver (Sovereign Alternative to Express — SIH 2026 | Wynit

SIH26119: Indigenous GPU-Accelerated Optimization Solver (Sovereign Alternative to Express — SIH 2026 | Wynit

What they'll ask you

11

Generated from the brief's own wording and the competition's published rules, never from a guess about what this ministry prefers.

Specific performance speedup targets compared to commercial solvers?
The brief never answers this, so a panel will. Whatever you decide, say it the same way twice.
Exact hardware specifications required for the GPU acceleration?
The brief never answers this, so a panel will. Whatever you decide, say it the same way twice.
Preferred programming language for building the solver core?
The brief never answers this, so a panel will. Whatever you decide, say it the same way twice.

8 morethe ones a panel gets to if there is time

Has any part of this been shown at a previous event, hackathon or college project?
The guidelines are explicit: your solution must not have appeared in any previous event or programme, of any sort. A recycled project is what a team under time pressure reaches for.
The brief asks for core solver engine. How would you build that?
Decoded from SIH26119 itself: Build algorithms from scratch for linear programming, mixed-integer linear programming, and quadratic programming without using existing open-source solver libraries. The brief asks for it by name.
The brief asks for numerical linear algebra. How would you build that?
Decoded from SIH26119 itself: Implement sparse matrix techniques and parallelization to handle large, complex industrial datasets. The brief asks for it by name.
The brief asks for api or cli interface. How would you build that?
Decoded from SIH26119 itself: Create a basic command-line interface or application programming interface to input problems and output solutions. The brief asks for it by name.
The brief asks for benchmark evaluation. How would you build that?
Decoded from SIH26119 itself: Test and compare solution quality and speed against standard benchmark libraries like MIPLIB or Netlib. The brief asks for it by name.
Why not use what already exists? Name the closest thing to this that is already running.
A team that has not named the alternative themselves is answering this for the first time in the room.
Which single thing will you demonstrate end to end, start to finish, with nothing skipped?
Ours, not a rule: a narrow thing that fully works survives questioning better than a broad thing that half works. If nobody on the team can name it, that is the finding.
Show me this working: the jury will check whether the engine is built from scratch without wrapping existing open-source solvers, whether it supports LP, MILP, and QP, and how it performs against standard benchmarks like MIPLIB or Netlib in terms of speed and numerical stability.
This is the evaluator read for your problem statement, decoded from the brief's own wording.

The government's original description (524 words)

• Background Almost every optimization problem in India's refining, petrochemical, power, logistics, manufacturing and planning sectors ultimately depends on a handful of foreign mathematical optimization solvers such as IBM ILOG CPLEX, Gurobi and FICO Xpress. These engines sit behind refinery scheduling, production planning, supply chain optimization, blending, energy management and many AI-driven decision-support systems. While they are extremely capable, they come with high recurring license costs, restrictive licensing models and limited visibility into the underlying optimization algorithms. Indian developers can formulate optimization problems, but they cannot inspect, modify or tailor the solver internals to suit strategic national requirements. Open-source alternatives such as COIN-OR CBC, HiGHS, GLPK and SCIP exist and have made significant progress, but they still lag behind commercial solvers for several classes of large-scale mixed-integer optimization problems and have not been developed, validated or optimized specifically for Indian industrial use cases. The real challenge is not building the modeling interface; it is developing a numerically robust optimization engine that consistently finds high-quality solutions for large, sparse and highly constrained industrial problems within practical computation times. • Description The objective is to develop a sovereign mathematical optimization solver core rather than a complete modeling environment. The solver should support Linear Programming (LP), Mixed-Integer Linear Programming (MILP) and Quadratic Programming (QP) as the initial focus, with a modular architecture that can later be extended to Mixed-Integer Quadratic Programming (MIQP), Nonlinear Programming (NLP) and Mixed-Integer Nonlinear Programming (MINLP). Core algorithms may include revised simplex and interior-point methods for continuous optimization, together with branch-and-bound, branch-and-cut, cutting planes, presolve, heuristics and advanced node selection strategies for mixed-integer problems. The solver should exploit sparse matrix techniques, efficient numerical linear algebra and multi-core parallelization, with GPU acceleration considered where it provides measurable benefits. The emphasis is on numerical stability, scalability and reliable convergence across large industrial optimization problems rather than on graphical interfaces or modelling tools. It shall not be built upon any existing open source solver library but shall be built from scratch from mathematical foundation. The scope is to solve optimization problems arising from refinery scheduling, crude blending, process optimization, production planning, logistics, power system dispatch, transportation and supply chain management. The benchmark is that the solver should consistently deliver optimal or near-optimal solutions for industrial-scale problems involving thousands to millions of variables and constraints, including highly degenerate models, ill-conditioned matrices and difficult mixed-integer formulations where weaker implementations exhibit excessive computation times or fail to converge. • Expected Solution A robust optimization engine with a basic application programming interface (API) or command-line interface is sufficient; a polished graphical user interface is not required. The solver should successfully solve standard benchmark problems from recognised optimization libraries such as MIPLIB, Netlib or Mittelmann benchmark sets, with solution quality and computational performance compared against at least one established commercial or open-source solver. A clear demonstration of numerical robustness should be provided by solving challenging large-scale optimization problems involving degeneracy, weak LP relaxations or ill-conditioned constraint matrices, where simpler implementations struggle to achieve reliable convergence or acceptable solution times. The resulting solver should provide a transparent, extensible and sovereign foundation for future Indian optimization software across industrial, scientific and strategic applications.