26 subfields. Real ones.

Plain-language primers on the core ideas of AI, machine learning, and deep learning — definitions, mental models, and how the pieces fit together. The on-ramp for newcomers.

Alignment research, interpretability, robustness, fairness, and responsible AI development.

LLM agents, planning, tool use, multi-agent systems, and autonomous workflows.

Real-world ML applications in healthcare, finance, climate, science, and industry.

Causal discovery, counterfactuals, treatment effect estimation, and causal representation learning.

Image recognition, object detection, segmentation, video analysis, and visual understanding.

Dataset construction, data quality, synthetic data, evaluation benchmarks, and data ops.

Neural network architectures, optimization theory, scaling laws, and foundational research.

Quantization, pruning, distillation, speculative decoding, and serving large models cheaply.

Privacy-preserving training, distributed optimization, and learning across edge devices.

Pretraining, scaling laws, instruction tuning, and the architecture of large general-purpose models.

Diffusion models, GANs, variational autoencoders, image/video/audio generation.

Message passing, graph transformers, link prediction, and learning on relational structure.

Dense retrieval, ranking, retrieval-augmented generation, and search system design.

Mechanistic interpretability, feature attribution, probing, and understanding what models learn.

Model deployment, monitoring, feature stores, experiment tracking, and ML pipelines.

Vision-language models, cross-modal alignment, video-language understanding, and grounded generation.

Language models, text generation, sentiment analysis, machine translation, and conversational AI.

Bayesian methods, variational inference, probabilistic programming, and uncertainty quantification.

Collaborative filtering, ranking, retrieval, and personalization at scale.

Policy optimization, multi-agent systems, reward modeling, and decision-making under uncertainty.

Robot learning, manipulation, navigation, sim-to-real transfer, and embodied agents.

ML for physics, chemistry, biology, materials, and the natural sciences.

Pretext tasks, contrastive learning, masked modeling, and representation learning without labels.

Speech recognition, text-to-speech, audio understanding, music generation, and voice cloning.

Sequence modeling, forecasting, anomaly detection, and dynamical systems.