Many companies are making efforts to apply AI technology but not many have succeeded in getting benefits by applying this technology to actual problems. Many examples, such as image classification, face recognition, language translation, etc., have shown that there are many things AI can do, but there are many engineering challenges that must be overcome to rip their benefits in real-world production environments. Trust is important before giving work to AI. AI systems are a new type of software systems that learn from data and that make software engineering activities, such as integration, testing, deployment, monitoring, and maintenance (retraining and upgrade of AI model), indispensable. Yet, there seems to be no systematic method yet for AI system engineering. In addition, it is necessary to consider building an AI platform for training, deploying, and operating models so that similar AI systems can be easily created. In this talk, I will introduce how we are trying to solve AI engineering problems at SK Hynix Inc. and share lessons learned from creating AI models and applying them to the field.