In the current era of educational digital transformation, there are structural problems in English teaching such as the lack of personalized guidance, insufficient utilization of data, and long evaluation and feedback times. Based on the data-driven learning theory, this paper integrates large language models and smart classroom technologies to design personalized English teaching and proposes an implementation path. According to the research findings, the promotion of smart classroom data is to reshape the teaching process into a cycle of language data occurrence - processing - analysis, and using "ad hoc data collection" to break the "data islands" caused by traditional education. A four-dimensional teaching design framework of "data diagnosis - precise teaching - dynamic evaluation - personalized push" is proposed, including three links: pre-class learning situation profiling, in-class human-computer cooperation, and post-class layered remediation. Through the data integration throughout the process, "teaching - learning - evaluation" can be achieved in consistency. It is believed that personalized teaching in smart classrooms is not about stacking technologies, but using data as an engine, transforming teaching decisions from "experience-based judgment" to "evidence-driven", and realizing large-scale individualized teaching.