Design a System to Recommend Local Restaurant Profiles
Overview
This question evaluates a data scientist's ability to design an end-to-end local recommender system, including competencies in data sourcing and labeling, feature engineering, candidate generation and ranking, real-time serving, latency and freshness trade-offs, and privacy-safe handling of location data.
Recommending Local Restaurant Pages in the News Feed
Context
Design a non-ads recommendation system within a large social media app to surface local restaurant business profiles in each user’s news feed. The goal is to maximize relevant engagement (e.g., profile clicks, saves, follows) while meeting latency, privacy, and user experience constraints.
Task
Describe an end-to-end design covering:
- Data sources and labels
- Key features (user, restaurant, and context)
- Model choices for candidate generation and ranking
- Real-time serving and ranking approach (including latency and freshness)
- Evaluation and iteration plan (offline and online)
Assume you can use standard logging, a feature store, and nearline streams. Prioritize local relevance and privacy-safe use of location data.
Solution
Login required