Design a mini ML system to recommend LinkedIn Learning courses to a user.
Product goal:
-
Recommend courses that help the user succeed in their job search and/or current role.
Available signals (examples):
-
Jobs the user applied to, jobs the user viewed/saved, jobs recommended to the user.
-
User profile (title, skills, seniority, industry), past course consumption, dwell/completion, search queries.
-
Course metadata (title, description, skills taught, difficulty, duration).
-
Job post text (description, requirements).
Requirements:
-
Propose an end-to-end architecture (data collection → feature generation → candidate generation → ranking → serving).
-
Explicitly describe how you would extract useful features from
job post text
(e.g., skills/requirements). Mention at least one non-LLM and one LLM-based approach.
-
Discuss training data/labels, offline and online evaluation metrics, and key failure modes (cold start, bias, feedback loops).
-
Briefly describe how you would incorporate business constraints (freshness, diversity, exploration, compliance).