With the rapid advancement of Artificial Intelligence technology, chip testing has evolved from a gatekeeper into a feedback mechanism to evolve the product spanning the entire product lifecycle (simulation, lab, mass production, and customer use). To address this paradigm shift, Gubo Technologies has integrated latest software technologies and data science to deliver innovative testing software solutions. By redefining testing workflows and establishing robust data governance frameworks, these solutions address three critical industry challenges: fragmented multi-stage data, post silicon validation efficiency bottlenecks, and the lack of systematic knowledge accumulation. This paper focuses on the validation of AMS (analog-mixed-signal) chips, detailing three key technological innovations: First, a structured chip knowledge repository integrates test case libraries, test IP libraries, and troubleshooting knowledges , achieving knowledge repository of testing expertise. Second, a cross-platform hardware abstraction layer ensures compatibility with mainstream testing instruments and custom hardware, enabling standardized automated testing frameworks. Finally, an LLM (Large Language Model)-based Agent architecture drives validation automation through natural language interaction. Furthermore, it establishes a self-reinforcing data ecosystem—failure modes identified during implementation are continuously fed back into the knowledge repository, generating high-quality training data for specialized AI models. This paves the way for autonomous fault diagnosis and intelligent solution recommendations. This report details a demo deployed in real-world scenarios. In an actual MCU chip post-silicon validation test, the AI Agent assists validation engineers by recommending test cases based on design specifications from R&D engineers, generating test plans upon confirmed test case by human (validation engineer) in loop and converting approved test plans into automated validation test programs. After automated test execution, it leverages the accumulated troubleshooting knowledge base to suggest potential causes to validation engineers. AI agent can also take the test results raw data to automatically aligned with design specifications, while the whole process of project team identifying the chip defects and confirmed root causes are systematically recorded by AI Agent. With AI Agent assistance, engineers can bypass tedious transactional tasks to concentrate on optimizing validation strategies and chip debugging, thereby significantly enhancing productivity.
This work not only validates the feasibility of intelligent testing systems in the semiconductor industry but also provides a reusable technical framework for the AI-driven transformation of testing methodologies. It propels chip testing beyond its traditional role as a "quality gatekeeper" toward becoming a "data-driven accelerator" for R&D, marking a leap forward in the evolution of testing technologies for the AI era.