SmartSolveAI CSAIL · MIT
MIT · CSAIL

AI-Guided
Algorithm Discovery
for Fast
Linear Algebra.

Numerical linear algebra is the computational foundation of nearly all modern scientific computing and AI, shaping performance from large-scale multiphysics simulations to the training and inference of today's largest AI models. Yet its design space — including algorithm classes, mixed-precision strategies, sparsity structures, memory layouts, hardware backends, and task scheduling — is far too large for any human team to explore comprehensively. SmartSolveAI investigates how multi-agent AI can help navigate this high-dimensional design space to discover specialized, high-performance algorithms that outperform today's hand-tuned systems.

Coming soon — a new agentic version is on the way. Stay tuned.
SmartSolve Designer

Describe your system.
Get a solver.

SmartSolve Designer generates a high-performance linear solver specialized to the problem at hand. It asks a short sequence of questions about your matrix, your right-hand sides, the baseline to beat and what would count as an improvement. Your answers are frozen into a run specification — and from there two agents, a Proposer and a Reviewer, collaborate until a candidate implementation beats a verified baseline. Any application in which Ax = b is a computational bottleneck is a potential target.

Wizard step 1 of 10, Matrix: which matrix do you want a faster solver for?
Step 01 of 10 · Matrix Which matrix do you want a faster solver for? Use the bundled example, generate one, or point to your own Matrix Market file. Full size ↗
  11. After the form
Requirements: SmartSolve Designer runs inside Claude Code and needs Julia 1.10 or newer; PETSc and an MPI C compiler are optional, and only required for the C backend.
Quick Start Tutorial↗ View on GitHub See more of our software→
Latest publications
Learning to Select Sparse Linear Solvers with Convolutional Neural Networks
Pados, Edelman, Lujan, Pickard, Tomé, Rackauckas · IEEE HPEC, 2026
Accelerating Linear Solves in Finite Element Simulations with Julia & AI
Lujan et al. · CHEFSI TST Meeting, April 2026 · Poster
All publications→
Latest talks
Automated Algorithm Selection Discovery via LLMs
Shah, Lujan, Alomairy · JuliaCon 2025 · Lightning Talk
Improving the HPC Experience: Did Julia Get It Right, or Will AI Hide the Problem?
Alan Edelman · WAMTA 2025 · Keynote
All talks→