01Deep Learning for InterviewsComplete roadmap for mastering the 11 core deep learning concepts tested in AI/ML interviews - backpropagation, activations, CNNs, RNNs, Transformers, normalization, training techniques, and more.02BackpropagationMaster backpropagation from first principles - chain rule, computational graphs, forward/backward pass, gradient derivations, vanishing/exploding gradients, and automatic differentiation for AI/ML interviews.03Activation FunctionsMaster every activation function tested in AI/ML interviews - sigmoid, tanh, ReLU, Leaky ReLU, PReLU, ELU, GELU, SiLU/Swish, Mish, and Softmax - with mathematical properties, gradient analysis, and architecture-specific selection criteria.04Convolutional Neural NetworksMaster CNNs for AI/ML interviews - convolution math, pooling, stride/padding, receptive fields, architecture evolution from LeNet to ConvNeXt, skip connections, depthwise separable convolutions, and transfer learning strategies.05RNNs and LSTMsMaster recurrent neural networks, vanishing gradients, LSTM gating mechanisms, GRUs, and bidirectional architectures - with interview-ready derivations and modern context on why Transformers replaced them.06Attention MechanismMaster attention from seq2seq bottleneck to multi-head self-attention - with QKV derivations, scaling intuition, attention pattern visualization, and interview-ready explanations.07Transformer ArchitectureMaster the complete Transformer architecture - encoder, decoder, positional encoding, pre-norm vs post-norm, BERT vs GPT, Flash Attention, KV cache, and modern efficiency techniques for AI interviews.08Normalization TechniquesMaster BatchNorm, LayerNorm, GroupNorm, InstanceNorm, and RMSNorm - with derivations, training vs inference behavior, and practical guidance on when to use which for AI interviews.09Training TechniquesWeight initialization, gradient clipping, mixed precision training, curriculum learning, knowledge distillation, and label smoothing - the toolkit that separates models that converge from models that collapse.10Distributed TrainingData parallelism, model parallelism, ZeRO, FSDP, DeepSpeed, and scaling laws - how to go from 1 GPU to 1000 GPUs and why Chinchilla changed everything.11Generative ModelsVAEs, GANs, diffusion models, and flow-based models - from ELBO derivations to DDPM score matching, with interview-ready explanations for every generative paradigm.12Deep Learning Interview Questions Bank65 deep learning interview questions with model answers and scoring rubrics - organized by screening, technical deep dive, senior/staff, company-tagged, and quick-fire rounds.
