AI-Driven Evolutionary Approach for Autonomous Quantum Algorithm Discovery

Advances in large language models (LLMs) have reached the point where they can now tackle grand challenge problems in science and technology. Google’s AlphaEvolve framework has demonstrated how ensembles of LLMs can be systematically orchestrated within an evolutionary AI pipeline to discover new algorithms, optimize complex systems, and make verifiable scientific discoveries. Building on this foundation, we developed our version of OpenEvolve (the open-source implementation of the AlphaEvolve) - an autonomous framework for quantum computing. Our OpenEvolve integrates LLMs with the hybrid computing platform CUDA-Q for accelerated quantum circuit simulation and employs an evolutionary strategy of mutation, evaluation, and selection to explore and optimize quantum algorithms. We demonstrate its capabilities through two representative applications: (1) optimizing Trotterization schemes for Hamiltonian time dynamics, and (2) discovering compact circuits for Hamiltonian evolution via efficient Pauli operator grouping. Our OpenEvolve leverages CUDA-Q for both Hamiltonian evolution and quantum circuit simulation. This work highlights the emerging role of LLM-driven autonomous systems as an important tool for advancing quantum computing.