ML interviews roadmap 2024

• Leetcode-style coding questions.
• Implement machine learning algorithms like SVM or some component of algorithms like backpropagation or convolution from scratch.
• Programming language-related questions in depth like about Python GIL or about C++ pointers.
• OOP-related theoretical and implementation questions.
• Typical SWE style system design interviews like design Instagram
• Machine learning system design interviews like a design a recommendation system.
• Machine learning theoretical questions like what is hinge loss or explain logistic regression or when could KL divergence be used.
• Deep learning theoretical questions like what's the difference between SGD and Adam, what is quantization in neural networks, how can you speed up inference of a deep learning model.
• Computer Vision theoretical questions like what's the difference between YOLO and FasterRCNN, what loss function could be used for image segmentation, or explain epipolar geometry.
• Natural Language Processing theoretical questions like how transformers are better than RNNs, what is bidirectional in BERT or what is the difference between stemming and lemmatization.
• Previous work, previous research paper, previous project-related questions.
• Take-home assignments are also all over the place from building a time series-based model to deploying a classification model as an endpoint to problems related to what their company is facing.
• Tools-related questions like Docker, Kubernetes, AWS, etc.
• Behavioral round interviews
• Math, statistics, and probability-based interviews like questions on Bayes theorem or on Bernoulli distribution or what is the rank of a matrix or differentiate something.