AI Methods for Quantum Circuit Optimization

Abstract

Full-day tutorial (TUT13) at IEEE Quantum Week 2025 (QCE25), co-presented with David Kremer and Victor Villar (IBM Quantum). Quantum circuit optimization is essential for getting the most out of current quantum hardware, and AI methods have emerged as a practical tool for circuit optimization and transpilation, striking a balance between output quality and computational effort. The first session gives an overview of the AI-powered transpiler passes available in Qiskit, with hands-on exercises applying them to produce optimized circuits and guidelines for using them effectively. The second session goes under the hood, covering the methods behind these passes and walking through training custom models for specific problems, including how to integrate them into Qiskit as transpiler passes. Hands-on material drew on the Qiskit Gym reinforcement-learning environments for quantum circuit synthesis. Aimed at beginner and intermediate users looking to leverage AI models for circuit optimization, as well as advanced practitioners interested in building their own AI-based optimization passes; foundational Python and quantum computing knowledge were assumed, with AI/ML and Qiskit experience helpful but not required.