A Systematic Review of Hybrid AI Frameworks Combining Circuit-Level Optimization and Machine Learning Approaches

This review focused on the integration of evolutionary, learning, and reinforcement techniques with circuit-level optimisation. The analysis indicated that hybrid AI excelled in adapting automation, fast convergence, and design flexibility beyond traditional heuristics and deterministic models. The bibliometrics showed a prominent and unrivalled rise in the design of intelligent hardware from 2015-2025 by Tsinghua University, the University of Chinese Academy of Sciences, and Harvard University, illustrating perceptible international cooperation in the field. The advancement of hybrid AI enabled self-predictive and adaptive learning electronic systems in FPGA optimisation, quantum circuit design, and neuromorphic architecture. With these hybrid systems, the models transition from a hardware design process of parameter tuning to one driven by data. The review provides significant contributions to the field by offering a hybrid AI systems taxonomy and synthesis while outlining the need for adaptively, real-time explainability, and lightweight models. Further work is also recommended for the development of AI-integrated EDA tools for automated circuit design. Overall, this study outlines the first comprehensive vision on hybrid AI–hardware co-design by marrying intelligent computation with engineering design for the development of smart, energy-efficient electronics.

1.      Agnesina, A., et al. (2020). VLSI Placement Parameter Optimization using Deep Reinforcement Learning. ICCAD paper / technical report. Retrieved from https://www.gtcad.gatech.edu/www/papers/agnesina-iccad20.pdf. gtcad.gatech.edu 2.      Attaoui, Y., Chentouf, M., & Ismaili, Z. E. A. (2024). Machine Learning in VLSI Design: A Comprehensive Review. Journal of Integrated Circuits and Systems. Retrieved from https://jics.org.br/ojs/index.php/JICS/article/download/826/500. jics.org.br+1 3.      Fernando, P. R. (2019). Genetic Algorithm Based Design and Optimization of VLSI. University/Technical Report. Retrieved from https://digitalcommons.usf.edu/cgi/viewcontent.cgi?article=2962&context=etd. digitalcommons.usf.edu 4.      Wang, Z. (2024). A Review on FPGA-Based Optimization and Applications of CNNs. CEUR Workshops. Retrieved from https://ceur-ws.org/Vol-3748/paper13.pdf. ceur-ws.org 5.      Karuppasamy, K., Puram, V., Johnson, S., & Thomas, J. P. (2024). A Comprehensive Review of Quantum Circuit Optimization: Current Trends and Future Directions. arXiv. Retrieved from https://arxiv.org/pdf/2408.08941. arXiv 6.      Paler, A. (2023). Machine Learning Optimization of Quantum Circuit Layouts. ACM Proceedings. Retrieved from https://dl.acm.org/doi/10.1145/3565271. dl.acm.org 7.      Riu, J., et al. (2025). Reinforcement Learning Based Quantum Circuit Optimization (ZX-diagram RL + GNNs). Quantum Journal. Retrieved from https://quantum-journal.org/papers/q-2025-05-28-1758/. Quantum 8.      “Machine Learning in VLSI layout design” (2024). Leveraging ML for enhanced automation in VLSI layout design. Journal/Report. Retrieved from https://urfjournals.org/open-access/leveraging-machine-learning-for-enhanced-automation-in-vlsi-layout-design-optimizing-performance-and-efficiency.pdf. urfjournals.org 9.      Research on FPGA ML accelerators (survey) — Darvish Rouhani, et al. (2024). A survey on FPGA-based accelerators for ML models. arXiv. Retrieved from https://arxiv.org/html/2412.15666v1. arXiv 10.   “Deep RL Global Router” — Enhanced DRL routing (2023). An Enhanced Deep Reinforcement Learning-Based Global Router for VLSI Design. Retrieved from https://www.researchgate.net/publication/370564725_An_Enhanced_Deep_Reinforcement_Learning-Based_Global_Router_for_VLSI_Design