I redesigned OLI's core learning workflows for 1,000+ students and modeled a shared system that translates course signals into personalized priorities for students and actionable interventions for instructors.
Telemetry is not guidance until the system can rank, explain, and route the next action.
Students saw progress but still had to decide what mattered. Instructors saw class-level signals but lacked a consistent path from risk detection to targeted intervention.
Students and instructors need different actions while relying on the same underlying mastery and activity data.
Low activity can mean confusion, absence, or simply an unfinished task; the UI must avoid false certainty.
Priority order affects what students study and where instructors spend limited intervention time.
The system normalizes learning signals, evaluates urgency and confidence, then changes its output according to who needs to act.
The student home prototype transforms multiple course signals into a small set of contextual next steps.
The instructor prototype preserves the path from cohort signal to student evidence and suggested intervention, reducing context switching.
When the signal set is incomplete, the system shows uncertainty and recommends observation instead of labeling a learner at risk.
Strong mastery and low participation are displayed as separate facts rather than collapsed into one misleading score.
Completed tasks, instructor overrides, and new attempts update the queue so recommendations do not persist beyond their relevance.
I worked with engineering to define the logic behind ranking, explanation, role permissions, and state transitions. Making those behaviors explicit reduced design-to-development iteration cycles by 40%.