This paper presents a new framework to harness sources of programming learning analytics at a Higher Education Institution and how it has been progressively adopted at the classroom level to improve personalized learning. This new platform, called PredictCS, automatically detects lower-performing or “at-risk” students in computer programming modules and automatically and adaptively sends them feedback. PredictCS embeds multiple predictive models by leveraging multi-modal learning analytics of student data, including student characteristics, prior academic history, logged interactions between students and online resources, and students' progress in programming laboratory work, and their progression from introductory to advanced CS courses. Predictions are generated every week during the semester's classes. In addition, students are flexible to opt-in to receive pseudo real-time personalized feedback, which permits them to be aware of their predicted course performance. The adaptive feedback ranges from programming suggestions from top- performers in the class to resources that are suitable to bridge their programing knowledge gaps.