← Meta Interview Insights

Meta·Software Engineer·Technical Phone Screen·Senior

Senior
Apr 2026

Summary

Went through a technical screen for an AI Research Scientist role at Meta. Heavy on ML fundamentals, the kind of stuff you think you know cold until someone's staring at you waiting for a precise answer.

Questions Asked (7)

Q1

Can you explain the bias-variance trade-off and how it affects model performance?

Technical Trade-offs
Author's notes

Knew this one but fumbled the articulation.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Start by defining bias and variance clearly, then explain the trade-off and its impact on model performance (underfitting vs. overfitting). Use a concrete example or a diagram to illustrate, and discuss strategies to balance the trade-off in practice.

Pro tip: Relate the trade-off to real-world engineering decisions, such as choosing model complexity or regularization, and mention how Meta's large-scale systems might require different trade-offs based on latency, accuracy, and resource constraints.

1. Define Bias and Variance

Clearly define bias as error from erroneous assumptions (underfitting) and variance as sensitivity to fluctuations in the training set (overfitting).

2. Explain the Trade-off

Describe how increasing model complexity decreases bias but increases variance, and vice versa, leading to a U-shaped test error curve.

3. Impact on Model Performance

Discuss how high bias leads to underfitting (poor on both train and test) and high variance leads to overfitting (good on train, poor on test).

4. Strategies to Balance

Mention techniques like cross-validation, regularization (L1/L2), ensemble methods (bagging/boosting), and early stopping to manage the trade-off.

5. Relate to Practical Scenarios

Connect to real-world examples, such as how Meta might prioritize low variance for stable predictions in production or accept higher bias for faster inference.

Key Points to Mention

  • Bias-variance decomposition of expected error
  • Underfitting vs. overfitting
  • Model complexity and its effect
  • Regularization techniques (L1, L2, dropout)
  • Ensemble methods (bagging reduces variance, boosting reduces bias)
  • Cross-validation for model selection

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

Q2

What are the differences between L1 and L2 regularization, and when would you choose one over the other? How does dropout fit into the regularization picture?

Technical Trade-offsAlgorithms & Data Structures
Author's notes

L1 vs L2 is pretty standard but I got a little tangled explaining why L1 produces sparse weights.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Start by defining L1 and L2 regularization mathematically and their effects on model weights, then compare their practical implications such as sparsity and robustness. Discuss when to choose each based on feature selection needs, computational constraints, and data characteristics. Finally, explain dropout as a different regularization technique that prevents co-adaptation and can be combined with L1/L2.

Pro tip: Mention that L1 is like a Laplace prior and L2 is like a Gaussian prior, showing Bayesian understanding, and note that dropout is particularly effective in deep neural networks where L1/L2 alone may not suffice.

1. Define L1 and L2

Explain that L1 adds the sum of absolute weights to the loss, promoting sparsity, while L2 adds the sum of squared weights, promoting small but non-zero weights.

2. Compare effects

Highlight that L1 can zero out irrelevant features (feature selection), while L2 distributes error across all weights, leading to smoother models and better handling of correlated features.

3. When to choose

Choose L1 when interpretability and feature selection are important; choose L2 when you have many small/medium effects or correlated features; consider Elastic Net (combination) for both.

4. Introduce dropout

Describe dropout as randomly dropping units during training, which prevents co-adaptation and acts as an ensemble method, often used in neural networks.

5. Integrate and conclude

Explain that dropout is complementary to L1/L2 and can be used together; L1/L2 are common in linear models, while dropout is specific to neural networks.

Key Points to Mention

  • L1 regularization leads to sparse solutions and can be used for feature selection.
  • L2 regularization penalizes large weights, improving generalization and handling multicollinearity.
  • L1 is robust to outliers but can be unstable with correlated features; L2 is more stable.
  • Dropout randomly deactivates neurons during training, reducing overfitting and co-adaptation.
  • Dropout can be seen as an ensemble of many sub-networks, similar to bagging.
  • Combining L1/L2 with dropout is common in deep learning for stronger regularization.

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

Q3

When would you use cross-entropy loss versus mean squared error, and what drives that choice?

Technical Trade-offs
Author's notes

Pretty comfortable here.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Start by defining the two loss functions and their underlying probabilistic assumptions, then explain that the choice depends on the nature of the target variable and the output activation. Use concrete examples to illustrate when each is appropriate, and mention practical considerations like gradient behavior and robustness to outliers.

Pro tip: Mention that cross-entropy is the natural choice for classification because it corresponds to maximum likelihood estimation for categorical distributions, while MSE is for regression under Gaussian noise assumptions. Also note that using MSE with sigmoid/softmax can lead to vanishing gradients, which is a common pitfall.

1. Define the losses and their assumptions

Briefly explain that cross-entropy measures the difference between two probability distributions and is used for classification, while MSE measures the average squared difference between predictions and targets and is used for regression.

2. Match loss to output type and activation

Explain that cross-entropy pairs with sigmoid (binary) or softmax (multiclass) outputs, while MSE typically pairs with linear outputs. Using the wrong combination can cause training issues like slow convergence.

3. Discuss probabilistic interpretations

Highlight that cross-entropy arises from maximum likelihood estimation for Bernoulli/Categorical distributions, while MSE arises from Gaussian noise assumptions. This justifies why each is used in its respective domain.

4. Consider practical factors

Mention that cross-entropy provides stronger gradients for classification, especially when predictions are wrong, while MSE is sensitive to outliers in regression. Also note that MSE can be used for classification but often performs worse.

5. Summarize with a clear rule of thumb

Conclude that the choice is driven by the problem type: cross-entropy for classification (discrete targets), MSE for regression (continuous targets), and sometimes other losses like MAE for robustness.

Key Points to Mention

  • Cross-entropy is for classification (binary/multiclass), MSE for regression.
  • Cross-entropy assumes a Bernoulli/Categorical distribution; MSE assumes Gaussian.
  • Using MSE with sigmoid/softmax can cause vanishing gradients.
  • Cross-entropy penalizes wrong predictions more strongly via log loss.
  • MSE is sensitive to outliers; MAE or Huber loss can be alternatives.
  • In practice, frameworks like PyTorch provide CrossEntropyLoss and MSELoss, and the choice affects convergence speed.

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

Q4

Walk me through the key evaluation metrics for classification and regression tasks. How do you decide which metric matters for a given problem?

Product Analytics & MetricsTechnical Trade-offs
Author's notes

Started with precision/recall/F1 and AUC for classification, RMSE and MAE for regression.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Start by clearly defining the most common evaluation metrics for classification and regression, explaining what each measures and when it is appropriate. Then, describe a systematic decision-making process that ties metric selection to the business problem, data characteristics, and model trade-offs. Use concrete examples to illustrate how the choice of metric impacts model behavior and outcomes.

Pro tip: Emphasize that the right metric aligns with the business objective and the cost of different error types—this shows you think beyond technical correctness. Mention that at companies like Meta, metrics are often tied to product goals (e.g., user engagement, false positive rates) and that you would validate metric choice with stakeholders.

1. Define Classification Metrics

List key classification metrics such as accuracy, precision, recall, F1-score, ROC-AUC, and PR-AUC, and briefly explain what each measures and its sensitivity to class imbalance.

2. Define Regression Metrics

List key regression metrics such as MSE, RMSE, MAE, MAPE, and R-squared, and explain how they penalize errors differently and their interpretability.

3. Map Metrics to Business Objectives

Explain how to align metric choice with the problem's goal: e.g., minimize false negatives in medical diagnosis (recall), minimize false positives in spam detection (precision), or minimize large errors in price prediction (RMSE).

4. Consider Data and Model Constraints

Discuss how class imbalance, outliers, and the need for interpretability influence metric selection, and how to use multiple metrics for a holistic view.

5. Iterate and Validate

Describe how to validate the chosen metric with stakeholders, monitor it post-deployment, and adjust if business priorities change.

Key Points to Mention

  • Accuracy is misleading for imbalanced datasets; use precision, recall, F1, or AUC instead.
  • Precision vs. recall trade-off depends on whether false positives or false negatives are more costly.
  • Regression metrics: MSE/RMSE penalize large errors heavily; MAE is more robust to outliers; R-squared measures variance explained.
  • Align metric with business KPI (e.g., click-through rate, conversion rate, customer lifetime value).
  • Use multiple metrics during development but select one primary metric for model selection and deployment.
  • Consider calibration and threshold tuning for classification metrics like precision and recall.

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

Q5

How do you set up train/validation/test splits, and when does cross-validation make more sense than a fixed split?

Algorithms & Data Structures
Author's notes

Said cross-validation is more useful when data is limited and you can't afford to hold out a big chunk.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Start by explaining the standard train/validation/test split and its purpose, then discuss when cross-validation is preferred, focusing on dataset size, variance, and computational trade-offs. Emphasize practical considerations like data leakage, stratification, and time-series splits.

Pro tip: Mention that for large datasets, a single validation set is often sufficient, but for small datasets, cross-validation provides more reliable performance estimates. Also, highlight that cross-validation is crucial for hyperparameter tuning to avoid overfitting to a single validation set.

1. Define the splits

Explain the typical 60/20/20 or 70/15/15 split for train/validation/test, and the role of each set: training for model fitting, validation for hyperparameter tuning and model selection, test for final unbiased evaluation.

2. Consider dataset size and variance

For small datasets, a fixed split may lead to high variance in performance estimates. Cross-validation (e.g., k-fold) uses all data for training and validation, providing a more robust estimate.

3. Account for data characteristics

Mention stratification for imbalanced classes, grouping for clustered data, and time-series splits to prevent data leakage. Cross-validation can be adapted (e.g., stratified k-fold, time-series split).

4. Evaluate computational cost

Cross-validation requires training k models, which can be computationally expensive. For large datasets or complex models, a fixed split may be more practical.

5. Decide based on goals

Use cross-validation when you need reliable performance estimates for model selection or hyperparameter tuning, especially with limited data. Use a fixed split when data is abundant and computational resources are limited.

Key Points to Mention

  • Data leakage: ensure test set is not used during training or validation, and preprocessing is fit on training data only.
  • Stratified sampling: preserve class distribution in splits for imbalanced datasets.
  • Time-series data: use temporal splits (e.g., train on past, validate on future) and avoid random shuffling.
  • Cross-validation variants: k-fold, stratified k-fold, leave-one-out, and nested cross-validation for hyperparameter tuning.
  • Bias-variance trade-off: cross-validation reduces variance in performance estimates but may increase bias if not done properly.
  • Computational efficiency: consider parallelization or using a single validation set for large datasets.

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

Q6

What are the main variants of gradient descent and how does learning rate scheduling factor into training stability?

Algorithms & Data StructuresTechnical Trade-offs
Author's notes

Covered SGD, mini-batch, Adam.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Start by categorizing gradient descent variants into three main types: batch, stochastic, and mini-batch, then discuss advanced optimizers like Momentum, RMSProp, and Adam. Explain how learning rate scheduling (e.g., step decay, cosine annealing, warm restarts) helps balance convergence speed and stability, and relate it to practical training scenarios.

Pro tip: Emphasize that the choice of variant and schedule depends on the problem scale and hardware constraints, and mention that adaptive methods like Adam often reduce the need for manual scheduling but can still benefit from it. Also, note that Meta often deals with large-scale recommendation and vision models, so highlighting experience with distributed training and learning rate warmup can set you apart.

1. Define gradient descent and its purpose

Briefly explain that gradient descent is an optimization algorithm to minimize a loss function by iteratively updating parameters in the direction of the negative gradient.

2. List main variants

Describe batch gradient descent (uses entire dataset), stochastic gradient descent (SGD, uses one sample), and mini-batch gradient descent (uses a subset). Mention that mini-batch is most common in practice.

3. Introduce advanced optimizers

Discuss variants like Momentum, Nesterov accelerated gradient, Adagrad, RMSProp, and Adam, explaining how they adapt learning rates or add momentum to improve convergence.

4. Explain learning rate scheduling

Describe common schedules: step decay, exponential decay, cosine annealing, and warm restarts. Explain how they adjust the learning rate over time to avoid overshooting minima and to escape saddle points.

5. Connect scheduling to training stability

Discuss how a high initial learning rate can cause divergence, while a low one can slow convergence. Scheduling helps maintain stability by reducing the learning rate as training progresses, and warmup can prevent early instability.

Key Points to Mention

  • Batch vs. stochastic vs. mini-batch gradient descent and their trade-offs in computation and convergence.
  • Advanced optimizers: Momentum, RMSProp, Adam, and their mechanisms (e.g., adaptive learning rates, moving averages).
  • Learning rate schedules: step decay, exponential decay, cosine annealing, and warm restarts.
  • The role of learning rate warmup in stabilizing early training, especially for large models.
  • How learning rate scheduling interacts with batch size and the concept of linear scaling rule.
  • Practical considerations: choosing optimizers and schedules based on model architecture, dataset size, and hardware.

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

Q7

How do you use learning curves to diagnose whether a model is underfitting or overfitting?

Root Cause AnalysisTechnical Trade-offs
Author's notes

This one I liked.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Start by defining learning curves and their axes, then explain how the gap between training and validation performance diagnoses underfitting vs. overfitting. Use a concrete example to illustrate the patterns and discuss how to act on each diagnosis.

Pro tip: Emphasize that learning curves are a diagnostic tool, not just a visualization—always pair them with a clear action plan (e.g., more data, regularization) to show you can drive model improvements.

1. Define learning curves

Explain that learning curves plot training and validation error (or accuracy) against training set size or training iterations. This sets the foundation for interpretation.

2. Identify underfitting pattern

Describe that underfitting occurs when both training and validation errors are high and converge to a similar, poor performance level. The model is too simple to capture the underlying pattern.

3. Identify overfitting pattern

Describe that overfitting occurs when training error is low but validation error is high, with a large gap between the two curves. The model memorizes training data but fails to generalize.

4. Prescribe remedies

For underfitting, suggest increasing model complexity, adding features, or reducing regularization. For overfitting, suggest getting more data, simplifying the model, or adding regularization.

5. Iterate and validate

Stress that learning curves should be re-plotted after each change to confirm the diagnosis and measure improvement, closing the loop on root cause analysis.

Key Points to Mention

  • Training vs. validation error gap as the key signal
  • Effect of training set size on curve convergence
  • Bias-variance tradeoff connection
  • Role of regularization and model complexity
  • Practical steps: more data, feature engineering, hyperparameter tuning
  • Avoiding common pitfalls like misinterpreting noisy curves

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