← Meta Interview Insights

Meta·Machine Learning Engineer·Technical Phone Screen·Senior

Senior
Jun 2026

Summary

Meta MLE interview focused heavily on a single ML system design problem around feature engineering for the News Feed ranking model. The follow-up about offline vs online metric gaps was where things got interesting and honestly where I felt most out of my depth.

Questions Asked (4)

Q1

An existing model predicts the probability a user likes a News Feed post. How would you add a feature tracking whether a user liked posts from a given author (or a specific post) in the last 6 months? Walk through feature engineering, how you'd serve it at inference time, and how you'd avoid label leakage during training.

System DesignTechnical Trade-offsData Modeling
Author's notes

This is where I spent most of the interview.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Start by defining the feature precisely: a time-decayed count of likes from the user to the author or post within the last 6 months. Then explain how to compute it offline with point-in-time correctness to avoid leakage, and how to serve it online with low latency using a feature store or precomputed aggregates. Finally, discuss trade-offs like storage, staleness, and cold-start handling.

Pro tip: Emphasize point-in-time correctness: when generating training data, only use interactions that occurred before the label timestamp. This is the most common source of leakage in temporal features and interviewers love to see you catch it.

1. Define the feature and its variants

Specify the exact feature: e.g., number of likes from user U to author A in the last 6 months, possibly with time decay or separate counts for post-level. Clarify how it will be used in the model (as a dense feature, bucketized, etc.).

2. Offline feature engineering with point-in-time correctness

Describe how to compute the feature from historical interaction logs. Use a time-windowed aggregation (e.g., sliding window over 6 months) and ensure that for each training example, only interactions before the label timestamp are included. Mention techniques like time-travel joins or snapshotting.

3. Online serving architecture

Explain how to serve the feature at inference time with low latency. Options: precompute aggregates in a key-value store (e.g., user-author like counts) and update incrementally, or use a feature store that supports real-time updates. Discuss trade-offs between freshness and cost.

4. Avoiding label leakage

Detail strategies to prevent leakage: strict temporal splits, excluding the current interaction from the count, and using only past data. Also mention the importance of consistent logic between offline and online computation to avoid training-serving skew.

5. Handle edge cases and evaluate impact

Discuss cold-start (new users/authors), sparsity, and how to handle missing values. Suggest evaluating the feature's impact via offline metrics and online A/B testing, and monitoring for drift.

Key Points to Mention

  • Point-in-time correctness: only use data available before the label timestamp to avoid leakage.
  • Time-decay or windowing: consider weighting recent interactions more heavily or using a fixed 6-month window.
  • Online-offline consistency: ensure the same aggregation logic is used in both training and serving to prevent training-serving skew.
  • Storage and latency trade-offs: precomputed aggregates vs. on-the-fly computation, and how to update them incrementally.
  • Cold-start and sparsity: fallback strategies for users or authors with no history.
  • Evaluation: offline metrics (AUC, log loss) and online A/B testing to measure feature impact.

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

Q2

How would you integrate this new feature into the existing model? Should it go in as a raw value, an embedding, or some kind of interaction term with existing features?

Technical Trade-offsProduct Analytics & Metrics
Author's notes

I went straight to interaction features because I thought that was the sophisticated answer.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Start by clarifying the feature's nature (e.g., categorical, continuous, text) and the model architecture. Then discuss trade-offs between raw value, embedding, and interaction terms, emphasizing the need to balance model complexity, interpretability, and performance. Finally, propose an empirical evaluation plan to test each option.

Pro tip: Always tie your recommendation to the specific product context and business metrics—at Meta, demonstrating that you consider both offline model metrics and online user impact sets you apart.

1. Understand the feature and model

Ask clarifying questions about the feature's type, cardinality, and relationship to existing features, as well as the current model architecture and constraints.

2. Evaluate representation options

Compare raw value, embedding, and interaction term approaches based on their ability to capture signal, computational cost, and risk of overfitting.

3. Consider model and business constraints

Assess how each option impacts latency, interpretability, and maintainability, and align with product goals and metrics.

4. Propose an experimental plan

Suggest offline experiments (e.g., A/B tests on a holdout set) to compare options, and define success metrics (e.g., AUC, CTR, revenue).

5. Recommend and iterate

Based on trade-offs and initial results, recommend a starting point and outline a plan to iterate, possibly combining approaches.

Key Points to Mention

  • Feature engineering best practices: when to use raw values vs. embeddings vs. interactions
  • Trade-offs between model complexity, interpretability, and performance
  • Impact on latency and scalability in production systems
  • Use of regularization to prevent overfitting with high-cardinality features
  • Offline evaluation metrics (e.g., AUC, log loss) and online metrics (e.g., CTR, engagement)
  • Iterative experimentation and monitoring post-deployment

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

Q3

How would you validate whether this feature actually helps? What offline metrics would you look at and how would you slice the analysis?

Product Analytics & MetricsA/B Testing & Experimentation
Author's notes

Talked through AUC and calibration, then slice analysis by content type and user activity level.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Start by clarifying the feature's goal and the primary success metric, then outline a validation plan that combines offline metrics with online A/B testing. Emphasize that offline metrics are proxies and should be validated online, and describe how you would slice the analysis to uncover heterogeneous effects.

Pro tip: Always tie offline metrics to the online business metric they aim to predict, and propose a guardrail metric to catch unintended consequences. Mention that you would pre-register the analysis plan to avoid p-hacking.

1. Clarify the feature's objective and success criteria

Ask what problem the feature solves and what the primary online metric is (e.g., CTR, engagement, revenue). Define what 'helps' means in measurable terms.

2. Select offline metrics that correlate with the online goal

Choose metrics like AUC, precision@k, recall, NDCG, or calibration that are relevant to the feature. Explain why they are proxies for the online metric.

3. Design offline evaluation with proper data splits

Use time-based splits to mimic production, avoid leakage, and evaluate on a holdout set. Consider counterfactual or off-policy evaluation if applicable.

4. Plan online validation via A/B test

Propose an A/B test with sufficient power, define primary and guardrail metrics, and specify duration and randomization unit.

5. Slice the analysis to understand heterogeneous effects

Segment by user demographics, behavior, device, geography, and other relevant dimensions. Look for consistent improvements and potential harms.

Key Points to Mention

  • Offline metrics are proxies; always validate online with an A/B test.
  • Choose offline metrics aligned with the product goal (e.g., ranking metrics for search, calibration for ads).
  • Use time-based splits to avoid temporal leakage and simulate real-world deployment.
  • Define guardrail metrics to monitor unintended consequences (e.g., user churn, latency).
  • Slice by user segments (new vs. existing, heavy vs. light users, demographics) to detect heterogeneous treatment effects.
  • Consider statistical power and multiple testing corrections when slicing.

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

Q4

Suppose your offline metrics improve significantly but the online A/B test shows no meaningful lift. What could explain that gap and how would you debug it?

A/B Testing & ExperimentationRoot Cause AnalysisProduct Analytics & Metrics
Author's notes

This was the hardest part and also the most interesting.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Acknowledge that offline-online gaps are common and often stem from differences in data, metrics, or user behavior. Structure your answer by first categorizing potential causes (data, metrics, model, experiment) and then outlining a systematic debugging process that involves validation, deeper analysis, and iteration.

Pro tip: Emphasize the importance of aligning offline evaluation with the online objective and using techniques like counterfactual logging or interleaving to bridge the gap. Show that you prioritize understanding the 'why' behind the metrics rather than just chasing numbers.

1. Validate the Experiment and Metrics

Check that the A/B test is properly set up: randomization, sample size, duration, and that the online metric is correctly defined and sensitive to changes. Ensure there are no bugs in logging or metric computation.

2. Compare Offline and Online Data Distributions

Investigate whether the offline training data matches the online serving data in terms of feature distributions, user segments, and temporal patterns. Look for covariate shift or concept drift.

3. Analyze Metric Alignment and Sensitivity

Examine if the offline metric is a good proxy for the online metric. Consider if the online metric is too coarse or if the treatment effect is diluted across segments. Check for novelty effects or metric saturation.

4. Inspect Model Behavior and Serving

Verify that the model is deployed correctly and performing as expected online. Check for training-serving skew, feature pipeline issues, or model staleness. Analyze if the model's predictions are actually influencing the user experience.

5. Deep Dive with Segment and Sensitivity Analysis

Break down the online results by user segments, time, and other dimensions to see if there are heterogeneous treatment effects. Perform power analysis to ensure the test had enough sensitivity to detect the expected lift.

Key Points to Mention

  • Training-serving skew and feature pipeline inconsistencies
  • Metric mismatch: offline metric not capturing online business objective
  • Experiment design issues: insufficient power, novelty effects, or interference
  • Data leakage or overfitting in offline evaluation
  • Heterogeneous treatment effects across user segments
  • Importance of guardrail metrics and long-term effects

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