GENIUS: an agentic AI framework for autonomous design and execution of simulation protocols

Abstract Predictive atomistic simulations have propelled materials discovery, yet routine setup and debugging still demand computer specialists. This know-how gap limits the use of Integrated Computational Materials Engineering (ICME), where state-of-the-art codes exist but remain cumbersome for non-experts. We address this bottleneck with GENIUS, an AI-agentic workflow that fuses a smart Quantum ESPRESSO knowledge graph with a tiered hierarchy of large language models supervised by a finite-state error-recovery machine. Here we show that GENIUS translates free-form human-generated prompts into Quantum ESPRESSO input files that pass early execution validation for ≈ 80% of 295 diverse benchmarks. Zero-shot generation succeeds for 14.2% of all prompts, and among cases that do not succeed initially, 76.3% are autonomously recovered by the automated error-handling loop, with the attempt-wise success rate decaying exponentially toward a 7% baseline. Compared with LLM-only baselines, GENIUS increases inference and computational efficiency and virtually eliminates hallucinations. The framework democratizes electronic-structure DFT simulations by intelligently automating protocol generation, validation, and repair, enabling large-scale screening and accelerating ICME design loops worldwide across academia and industry.