Smart contracts are a key technology in decentralized finance (DeFi) today. However, news of significant financial losses arising from contract vulnerabilities is far from infrequent. Indeed, we often hear about multi-million-dollar losses due to an error in a single line of code or in some aspect of the logic design. This underscores how the need for rigor in smart contract development goes far beyond that of generic software. Fortunately, outside of the costly process of manual auditing, there are now various rule/fuzzing-based automated scanning tools available that allow developers to conduct programming checks more efficiently and inexpensively. Training AI models on a large volume of contracts and developing models based on Transformer technology has enabled us to develop an AI vulnerability detection tool featuring both high accuracy and speed.