#deep learning

9 deep learning interview questions, asked at Amazon, Google, Netflix and 5 more companies. Last asked Oct 2026.

+7
ML fundamentals deep dive for ML rolesAI/ML designAfter a short walk through your resume, expect a fast run of questions on machine learning basics, usually centred on your own area (language models, vision, recommendations). For…
Last asked Oct 2026
+2
Transformer and attention fundamentalsAI/ML designA conceptual deep dive on transformers, asked of ML candidates. Explain how a transformer works from the inside, deriving the parts on a whiteboard rather than naming them. Expect…
Last asked Sep 2026
Softmax cross-entropy forward and backward, then a training loopCodingImplement from scratch the forward and backward pass of a softmax layer followed by cross-entropy loss. You may use NumPy but not automatic differentiation. Then use it to train a…
Last asked Sep 2026
VAEs: the ELBO, the KL term and reparameterizationAI/ML designDefine a variational autoencoder and derive its evidence lower bound (ELBO). Explain why the objective contains a KL divergence term and what that term does, explain the…
Last asked Aug 2026
Dataset size, generalization and U-Net skip connectionsAI/ML designA set of machine learning fundamentals questions for a vision role: how training set size affects generalization, how you would plan modeling end to end, how the loss function…
Last asked Mar 2026
2D convolution forward pass in NumPy, with parameter countsCodingImplement the forward pass of a 2D convolution from scratch, with no deep-learning library and no built-in convolution. Before coding, answer how many parameters the layer has and…
Last asked Dec 2025
Write a PyTorch training loopCodingWrite a training loop in PyTorch. You are given a model, a DataLoader that yields (inputs, targets) batches, an optimizer and a loss function. Train for a number of epochs and…
Last asked Dec 2025
Implement a decoder-only GPT modelCodingWrite a small decoder-only transformer language model (GPT-style) from scratch, for next-token prediction. Use PyTorch or another framework for tensors, but write the layers…
Last asked Nov 2025
Design a multimodal neural networkAI/ML designDesign a neural network that takes text and images together and uses them for retrieval (find the images for a text query, or the text for an image) and for classification (for…
Last asked Sep 2025