Meta·Machine Learning Engineer·Onsite - System Design / Architecture·Senior

Senior

Jul 2026

Summary

ML system design round at Meta for an MLE role. The whole thing was a single deep-dive into recommendation systems, and they really wanted you to go end-to-end, not just sketch a two-tower model and call it done.

Questions Asked (1)

Design a large-scale recommendation system for a feed or short-video product. Walk through your business objectives and success metrics, how you'd generate candidates, the ranking model and its features, training data and labels, serving architecture and latency constraints, A/B testing strategy, feedback loops, and cold-start handling.

System DesignA/B Testing & ExperimentationTechnical Trade-offs

Author's notes

This is basically the whole interview in one question.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Start by clarifying the product goals and success metrics, then walk through the ML pipeline from candidate generation to serving, emphasizing trade-offs and scalability. Structure your answer around the user journey and iterate on the design based on feedback and experimentation.

Pro tip: Focus on the two-stage ranking (candidate generation + ranking) and explain how you balance relevance with diversity and freshness. Mention how you'd handle cold-start by leveraging content-based features and explore/exploit strategies.

1. Define Objectives and Metrics

Clarify business goals (e.g., engagement, retention) and translate them into measurable ML metrics (e.g., CTR, watch time). Discuss trade-offs between short-term and long-term metrics.

2. Candidate Generation

Describe multiple candidate sources (e.g., collaborative filtering, content-based, trending) and how to combine them. Mention scalability and coverage considerations.

3. Ranking Model and Features

Explain the ranking model (e.g., deep neural network) and key features (user, item, context, cross features). Discuss how to handle feature engineering and model training.

4. Serving Architecture and Latency

Outline a scalable serving architecture (e.g., microservices, caching, pre-computation) and latency constraints (e.g., <200ms). Discuss trade-offs between model complexity and latency.

5. Experimentation and Feedback Loops

Describe A/B testing methodology, metrics, and how to incorporate feedback (e.g., online learning, retraining). Address cold-start with exploration strategies and content features.

Key Points to Mention

Two-stage architecture: candidate generation and ranking
Feature engineering: user, item, context, and cross features
Latency constraints and serving optimizations (caching, model distillation)
A/B testing framework and guardrail metrics
Cold-start solutions: content-based features, explore/exploit, meta-learning
Feedback loops: online learning, retraining cadence, and bias correction

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