LinkedIn·Machine Learning Engineer·Technical Phone Screen·Senior

Senior

Jun 2026

Summary

LinkedIn ML engineer interview, one question about building a location-aware job recommendation feature. Pretty thin on details but the core problem was interesting enough to think through.

Questions Asked (1)

How would you design a system to show job postings to users in the same city, personalized based on their past search behavior?

System DesignProduct Sense & IdeationTechnical Trade-offs

Author's notes

I went straight to the retrieval side, filtering by geo and then ranking with a learned model on search history.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Start by clarifying the requirements and constraints, then outline a high-level system architecture that includes data collection, candidate generation, ranking, and serving. Focus on how to combine geographic filtering with personalized ranking using past search behavior, and discuss trade-offs between relevance, freshness, and scalability.

Pro tip: Emphasize the importance of online-offline consistency and how you would handle cold-start users by leveraging location-based popularity or trending jobs. Also, mention the need for A/B testing and feedback loops to continuously improve personalization.

1. Clarify Requirements and Scope

Ask questions to understand scale, latency requirements, data availability, and success metrics. Define what 'same city' means (e.g., within a radius) and how past search behavior is tracked.

2. High-Level Architecture

Propose a system with offline and online components: data ingestion, feature engineering, model training, and a serving layer. Include a geo-filtering service and a personalization engine.

3. Data and Feature Engineering

Describe data sources: user profiles, search logs, job postings. Outline features like user location, search queries, clicked jobs, and job attributes. Discuss how to handle sparsity and recency.

4. Modeling and Ranking

Explain candidate generation (e.g., geo-based retrieval) and ranking (e.g., learning-to-rank with personalization). Mention two-tower models or gradient boosted trees, and how to incorporate location as a feature.

5. Serving, Evaluation, and Iteration

Discuss low-latency serving, caching, and fallback strategies. Cover offline metrics (NDCG, recall) and online metrics (CTR, apply rate). Mention A/B testing and monitoring for drift.

Key Points to Mention

Two-stage retrieval and ranking: candidate generation via geo-filtering, then personalized ranking.
Feature engineering: user location, search history, job attributes, and contextual signals.
Handling cold-start: use location-based popularity or trending jobs for new users.
Scalability: sharding by city, caching, and approximate nearest neighbor for geo queries.
Evaluation: offline metrics (precision@k, NDCG) and online A/B tests (CTR, application rate).
Trade-offs: relevance vs. freshness, personalization vs. diversity, and latency vs. model complexity.

AI-generated suggestions, not part of the candidate's original notes. May be inaccurate — verify before relying on them.