The continuous increase in complexity and scale of cloud platforms, together with the evolving diversity of services, are forcing AIOps to face platform characteristics that can significant impact the availability of cloud services. This talk will start from how to understand the running behavior of cloud platforms and services in terms of resource usage and request execution, and then address issues of AIOps with a focus on how the field of operation engineering can become more efficient through streaming processing, active learning, and knowledge-enhanced dialogue generation.