Machine Learning Interview Preparation Checklist (MAANG Level)

CLASSES
https://www.youtube.com/playlist?list=PLKnIA16_Rmvbr7zKYQuBfsVkjoLcJgxHH

STATSQUEST
https://www.youtube.com/playlist?list=PLblh5JKOoLUICTaGLRoHQDuF_7q2GfuJF

Andrew ng
https://www.youtube.com/playlist?list=PLoROMvodv4rMiGQp3WXShtMGgzqpfVfbU
https://www.youtube.com/playlist?list=PLkDaE6sCZn6FNC6YRfRQc_FbeQrF8BwGI

STANFORD
https://www.youtube.com/playlist?list=PLoROMvodv4rNyWOpJg_Yh4NSqI4Z4vOYy
https://www.youtube.com/playlist?list=PLoROMvodv4rNH7qL6-efu_q2_bPuy0adh

Krish naik
https://www.youtube.com/playlist?list=PLTDARY42LDV7WGmlzZtY-w9pemyPrKNUZ
https://www.youtube.com/playlist?list=PLZoTAELRMXVPBTrWtJkn3wWQxZkmTXGwe

MIT
https://openlearninglibrary.mit.edu/courses/course-v1:MITx+6.036+1T2019/course/

REVISION
https://www.youtube.com/playlist?list=PLfFghEzKVmjsNtIRwErklMAN8nJmebB0I
SIMPLILEARN
https://www.youtube.com/playlist?list=PLEiEAq2VkUULYYgj13YHUWmRePqiu8Ddy

NOTES
https://www.geeksforgeeks.org/100-days-of-machine-learning/?ref=shm

QUESTIONS
https://www.geeksforgeeks.org/machine-learning-interview-questions/?ref=shm

1. Machine Learning Fundamentals

✅ Topics:
 Supervised vs. Unsupervised Learning
 Types of Machine Learning Algorithms
 Bias-Variance Tradeoff
 Curse of Dimensionality
 Overfitting & Underfitting
 Model Interpretability (SHAP, LIME, Feature Importance)
✅ Question Types:
 Explain key ML concepts theoretically
 Given a dataset, which ML approach would you choose and why?
 Coding problems (train a model using Scikit-Learn, tune hyperpar
 ameters)
  Solve a bias-variance tradeoff numerical problem
 Compare overfitting vs. underfitting in a given scenario

✅ Depth: Intermediate to Advanced

✅ Common Pitfalls: Misidentifying when a model is overfitting or underfitting,
misunderstanding interpretability methods

Fundamentals of Machine Learning

 Supervised Learning (Regression, Classification)
 Unsupervised Learning (Clustering, Dimensionality Reduction)
 Semi-supervised and Self-supervised Learning
 Reinforcement Learning Basics
II. Statistical Foundations
 Probability Distributions & Bayes Theorem
 Hypothesis Testing & Confidence Intervals
 Maximum Likelihood Estimation (MLE) & MAP Estimation
 Bias-Variance Tradeoff
 Overfitting, Underfitting, and Regularization
III. Model Evaluation & Metrics
 Precision, Recall, F1-Score, ROC-AUC, Log-loss
 Cross-validation Techniques (k-fold, Stratified, LOOCV)
 Bootstrap Sampling
 A/B Testing
IV. Feature Engineering & Data Preprocessing
 Handling Missing Data, Outliers
 Feature Scaling, Encoding Categorical Variables
 Feature Selection & Dimensionality Reduction (PCA, LDA, t-SNE)
 Feature Engineering for Text & Time-Series Data

2. Linear Regression
✅ Topics:
 Ordinary Least Squares (OLS)
 Assumptions of Linear Regression
 Multicollinearity, Autocorrelation
 R-squared & Adjusted R-squared
 Regularization (Ridge & Lasso)
✅ Question Types:
 Theoretical: Explain assumptions, impact of violating assumptions
 Numerical: Compute OLS manually, interpret coefficients
 Coding: Fit a regression model, handle multicollinearity, use L1/L2 regularization
 Application: How does multicollinearity affect feature importance?
✅ Depth: Advanced
✅ Common Pitfalls: Assuming linearity in all data, incorrect handling of categorical
variables

3. Logistic Regression & Classification Metrics

✅ Topics:
  Sigmoid Function & Decision Boundary
 Assumptions of Logistic Regression
 ROC Curve, AUC, Precision-Recall, F1-score
 Handling Imbalanced Data (SMOTE, Class Weights)
✅ Question Types:
 Explain log odds and probability interpretation
 Calculate precision, recall, F1-score for given confusion matrix
 Coding: Implement logistic regression, tune hyperparameters, handle class
imbalance
✅ Depth: Advanced
✅ Common Pitfalls: Incorrect metric choice for imbalanced classification tasks

4. Decision Trees & Ensemble Methods

✅ Topics:
 Gini Index, Entropy, Information Gain
 Overfitting in Decision Trees, Pruning
 Bagging, Boosting (Random Forest, XGBoost, LightGBM, CatBoost)
 Feature Importance & SHAP Values
✅ Question Types:
 Explain Gini vs. Entropy
 Manually compute splits for a decision tree
 Implement and optimize Random Forest/XGBoost in Python
✅ Depth: Advanced
✅ Common Pitfalls: Ignoring feature selection in tree-based models

5. Support Vector Machines (SVM)

✅ Topics:
 Hyperplanes, Margin Maximization
 Kernel Trick (Polynomial, RBF)
 SVM vs. Logistic Regression
✅ Question Types:
 Explain kernel trick and why it’s needed
 Compute margin width given support vectors
 Coding: Implement SVM with different kernels
✅ Depth: Intermediate to Advanced
✅ Common Pitfalls: Misunderstanding kernel functions and their impact on computational
cost

6. k-Nearest Neighbors (kNN) & Clustering

✅ Topics:
 Distance Metrics (Euclidean, Manhattan, Cosine)
 Choosing the Right k
 DBSCAN, Hierarchical Clustering
 Silhouette Score, Elbow Method
✅ Question Types:
 Compute distance metrics manually
 Apply kNN for classification in Python
  Explain when k-means fails and how to fix it
✅ Depth: Intermediate
✅ Common Pitfalls: Not normalizing data before applying kNN

7. Principal Component Analysis (PCA) & Dimensionality Reduction

✅ Topics:
 Eigenvalues & Eigenvectors
 Explained Variance Ratio
 When to use PCA vs. t-SNE vs. UMAP
✅ Question Types:
 Manually compute PCA transformation
 Implement PCA in Python & interpret explained variance
 Compare PCA vs. t-SNE in different scenarios
✅ Depth: Advanced
✅ Common Pitfalls: Applying PCA blindly without considering feature interpretability

8. Probability & Bayesian Methods in ML

✅ Topics:
 Naïve Bayes Classifier
 Maximum Likelihood Estimation (MLE)
 Bayesian Inference & Bayesian Optimization
✅ Question Types:
 Explain Naïve Bayes assumptions
 Compute posterior probability manually
 Implement Naïve Bayes in Python
✅ Depth: Intermediate to Advanced
✅ Common Pitfalls: Misunderstanding conditional independence assumption

9. Neural Networks & Deep Learning Basics

✅ Topics:
 Perceptron, Activation Functions
 Backpropagation, Gradient Descent (SGD, Adam)
 CNNs vs. RNNs vs. Transformers
✅ Question Types:
 Explain forward & backward propagation
 Implement a simple MLP in TensorFlow/PyTorch
 Compute gradient updates manually
✅ Depth: Advanced
✅ Common Pitfalls: Choosing incorrect learning rate, overfitting due to excessive
parameters

10. Model Evaluation & Hyperparameter Tuning

✅ Topics:
 Cross-Validation (K-Fold, Stratified)
 Grid Search vs. Random Search vs. Bayesian Optimization
 Performance Metrics for Regression & Classification
✅ Question Types:
  Compute k-fold CV manually
 Implement hyperparameter tuning in Python
 Explain when to use F1-score vs. ROC-AUC
✅ Depth: Advanced
✅ Common Pitfalls: Ignoring cross-validation, misunderstanding metric trade-offs

11. Feature Engineering & Data Preprocessing

✅ Topics:
 One-Hot Encoding vs. Label Encoding
 Handling Missing Values (Mean Imputation, KNN Imputation)
 Feature Scaling (Min-Max, Standardization)
✅ Question Types:
 Coding: Preprocess raw dataset for ML model
 Theoretical: How does feature scaling affect distance-based models?
✅ Depth: Advanced
✅ Common Pitfalls: Incorrect encoding of categorical variables leading to data leakage

Strategies & Tips for Success

✅ Prioritize Depth Over Breadth: MAANG companies expect in-depth knowledge. Master
fundamental concepts deeply rather than superficially covering everything.
✅ Practice on Real-world Datasets: Kaggle competitions & real-world projects are the
best way to gain hands-on experience.
✅ Improve Speed in Manual Calculations: Expect questions where you must calculate
probabilities, entropy, or regression coefficients manually.
✅ Optimize Code for Efficiency: Focus on writing clean, efficient Python code using
NumPy, Pandas, and Scikit-Learn.
✅ Mock Interviews & Whiteboarding: Simulate real interview conditions by solving
problems on a whiteboard or explaining your thought process aloud.
✅ Stay Updated with Research: Read ML papers, follow top researchers, and stay up-to-
date with advancements in ML techniques.

Advanced Machine Learning Curriculum for MAANG Interviews

1. Core Topics
A structured breakdown of key machine learning concepts with Python:
I.
V. Machine Learning Algorithms
 Linear & Logistic Regression
 Decision Trees & Random Forests
 Support Vector Machines (SVMs)
 k-Nearest Neighbors (k-NN)
 Gradient Boosting (XGBoost, LightGBM, CatBoost)
 Neural Networks & Deep Learning Fundamentals
 Bayesian Methods & Probabilistic Graphical Models
VI. Advanced Topics
 Generative Models (GANs, VAEs)
  Transformer-based Architectures (BERT, GPT)
 Reinforcement Learning (DQN, PPO, Actor-Critic)
 AutoML & Hyperparameter Optimization
 Explainable AI (SHAP, LIME, Feature Importance)

2. Question Types
For each topic, expect questions in the following categories:
 Theoretical Questions (Explain concepts, trade-offs, comparisons)
 Conceptual Problem Solving (Derive equations, intuitive explanations)
 Coding Challenges (Implement models from scratch, optimize algorithms)
 Debugging Questions (Identify issues in ML models and pipelines)
 Design Patterns (ML system design, model scalability)
 Simulation-based Questions (Analyze hypothetical scenarios)
 Best Practices & Trade-offs (Optimize performance, handle real-world challenges)

3. Depth & Real-World Application Examples

Each topic requires different levels of understanding:
 Basic: Implementing models in Python using sklearn, numpy, pandas
 Intermediate: Hyperparameter tuning, improving model performance
 Advanced: End-to-end ML systems, deployment, distributed training
Examples:
 Bias-Variance Tradeoff → Tuning regularization in Ridge Regression
 Model Evaluation → Choosing metrics for imbalanced datasets
 Feature Engineering → Creating domain-specific features for NLP

4. Common Pitfalls
 Confusing correlation with causation
 Over-reliance on accuracy as a metric
 Ignoring data leakage
 Misinterpreting bias-variance tradeoff
 Improper handling of class imbalance
 Poor feature selection leading to model degradation

5. Advanced Applications
 NLP: Transformer-based models, sentiment analysis
 Computer Vision: CNN architectures, object detection
 Time Series: Forecasting with LSTMs, Prophet
 Recommendation Systems: Collaborative filtering, matrix factorization
 AutoML: Automating model selection & hyperparameter tuning
 ML in Production: MLOps, model deployment, monitoring drift

6. Practice & Resources

 Books:
o "Pattern Recognition and Machine Learning" - Christopher Bishop
o "The Elements of Statistical Learning" - Hastie, Tibshirani, Friedman
o "Deep Learning" - Ian Goodfellow, Yoshua Bengio, Aaron Courville
 o "Hands-On Machine Learning with Scikit-Learn, Keras & TensorFlow" -
Géron
 Courses:
o Andrew Ng’s ML & DL Courses (Coursera)
o Stanford CS229 (ML), CS231n (CV), CS224n (NLP)
 Coding Platforms:
o LeetCode (ML & SQL problems)
o Kaggle (Competitions & Datasets)
o Google Colab (Python Notebooks)
 Research Papers:
o "Attention Is All You Need" (Transformers)
o "ImageNet Classification with Deep Convolutional Networks" (AlexNet)
o "Auto-Encoding Variational Bayes" (VAEs)

7. Interview Strategy
Structuring Answers
 Theoretical: Define → Explain → Example → Use Cases
 Coding: Clarify → Plan → Implement → Optimize → Test
 ML System Design: Problem Breakdown → Data Pipeline → Model Choice →
Deployment Plan → Scaling
Common Tricks & Best Practices
 Think aloud to demonstrate reasoning
 Trade-offs: When to use complex vs. simple models
 Debugging: Find model errors via loss curves, feature importance
 Edge cases: Handle extreme values, unseen categories
Time Management
 Theoretical Qs: Answer in 2-3 min
 Coding Qs: Allocate time (15 min for approach, 30 min for coding, 15 min for
debugging & optimization)

1. Types of Questions in MAANG ML Interviews

1.1 Theoretical Questions
 What is the difference between supervised, unsupervised, and reinforcement
learning?
 Explain overfitting and underfitting. How can you prevent them?
 What is bias-variance tradeoff?
 What are precision, recall, F1-score, and accuracy?
 Explain feature selection techniques.
 What are different types of distance metrics (Euclidean, Manhattan, Cosine
Similarity)?
 Explain different types of activation functions in neural networks.
 What are loss functions and how do you choose them?
1.2 Conceptual Questions
 How do decision trees work? What are entropy and information gain?
 When should you use a support vector machine vs. a neural network?
 How does gradient descent work, and what are different variants like SGD, Adam,
RMSprop?
  How do convolutional neural networks (CNNs) work?
 Explain Principal Component Analysis (PCA) and when to use it.
 What is transfer learning, and why is it useful?
1.3 Numerical Problems
 Given a dataset, manually calculate Gini impurity and entropy for a decision tree.
 Compute precision, recall, and F1-score from a confusion matrix.
 Calculate the number of parameters in a neural network given layer dimensions.
 Compute eigenvalues and eigenvectors for PCA manually.
 Given a probability distribution, calculate expected value and variance.
1.4 Coding-Based Questions
 Implement logistic regression from scratch.
 Write a function to calculate the Euclidean distance between two points.
 Implement k-means clustering from scratch.
 Build a simple neural network using NumPy only (no deep learning libraries).
 Write code to preprocess missing values in a dataset.
 Implement gradient descent for linear regression.
1.5 Simulation-Based Questions
 How would you detect fraudulent transactions using machine learning?
 Design a recommendation system for an e-commerce website.
 How would you use ML to classify spam emails?
 Predict stock prices using machine learning models.
 Apply clustering for customer segmentation in a marketing campaign.
1.6 Design-Based Questions
 How would you design a system that can predict demand for an online grocery
store?
 How would you build a scalable recommendation system for Netflix?
 How do you design an ML pipeline for real-time fraud detection?
 How do you handle concept drift in a production ML model?
1.7 Debugging Questions
 Why is your deep learning model not converging?
 How do you fix a random forest that is overfitting?
 Your logistic regression model has low accuracy; what steps would you take?
 Why is your gradient descent algorithm getting stuck in local minima?
1.8 Pattern-Based Questions
 Identify edge cases for an anomaly detection model.
 How do you handle unbalanced datasets in classification?
 What are common pitfalls in hyperparameter tuning?
 How do you choose the right ML model for different types of data?
1.9 Optimization Problems
 How would you reduce the computational cost of training deep learning models?
 How can you optimize hyperparameters efficiently?
 How do you speed up inference time for a large-scale ML model?
 How do you optimize memory usage when working with large datasets?
1.10 Application-Based Questions
 How is ML used in self-driving cars?
 How does YouTube recommend videos?
 How does Google Search use ML?
 How does ML improve cybersecurity threat detection?
2. Common Pitfalls in ML Interviews
 Failing to explain why a specific model is chosen over others.
 Ignoring feature engineering and assuming raw data is good enough.
 Not considering edge cases and failure scenarios in real-world applications.
 Struggling with debugging when a model doesn’t perform well.
 Over-reliance on deep learning when simpler models work better.
 Not understanding time complexity and scalability of ML algorithms.
 Poorly explaining trade-offs in model selection.

3. Best Resources for Mastering ML

3.1 Books
 "Pattern Recognition and Machine Learning" – Christopher Bishop
 "The Elements of Statistical Learning" – Hastie, Tibshirani, Friedman
 "Hands-On Machine Learning with Scikit-Learn, Keras, and TensorFlow" – Aurélien
Géron
 "Deep Learning" – Ian Goodfellow
3.2 Videos & Courses
 Andrew Ng's Machine Learning Course (Coursera) – Link
 Fast.ai Practical Deep Learning for Coders – Link
 MIT OpenCourseWare on ML – Link
3.3 PPTs and Notes
 Stanford CS229 ML Notes – Link
 MIT Deep Learning Lecture Slides – Link
3.4 Question Banks
 Machine Learning Interviews Book by Chip Huyen – Link
 Interview Query – Link
3.5 Coding Platforms
 LeetCode (ML Section) – Link
 Kaggle ML Competitions – Link
 HackerRank ML Challenges – Link
3.6 Research Papers
 "Attention is All You Need" – Vaswani et al. (Transformers)
 "Batch Normalization: Accelerating Deep Network Training"
 "Gradient-Based Optimization of Hyperparameters"

4. Practice Strategy
4.1 Learning Approach
 Start with theoretical understanding: Watch Andrew Ng’s ML course.
 Follow with hands-on coding: Implement algorithms in Python from scratch.
 Work on real-world datasets: Use Kaggle and UCI ML Repository.
4.2 Balancing Theory and Coding
 Morning: Read ML concepts from books/resources.
 Afternoon: Solve coding problems on LeetCode/Kaggle.
 Evening: Work on ML projects or case studies.
4.3 Handling Edge Cases & Complexity
 Identify corner cases in classification problems (e.g., class imbalance).
  Optimize models for both accuracy and efficiency.
 Learn parallelization and distributed computing (Spark, Dask).
4.4 Preparing for Follow-Up Questions
 Be ready to explain how to improve an ML model step-by-step.
 Justify why you selected a model based on trade-offs.
 Discuss challenges faced in real-world ML deployment.

Final Words
By following this structured approach, you'll be well-prepared for any ML-related interview
question at MAANG companies. Focus on deep conceptual understanding, hands-on
coding, and real-world application to gain confidence in handling diverse ML problems.
Would you like additional mock interview questions tailored to your specific weak areas? 😊