Learn AI by Rajul Babel - Visual Interactive Guide to AI Internals

Learn AI is a free interactive guide by Rajul Babel that teaches Artificial Intelligence from first principles to production systems. Every concept is shown with real numbers, real formulas, and progressive visual diagrams - no watered-down explanations.

What you will learn

• Neural Networks - mechanics, neurons, layers, weights, activation, the forward pass.
• Learning & Backprop - loss functions, gradients, gradient descent, deep backprop.
• Linear Algebra for Deep Learning - vectors, dot products, matrices, activation.
• Training Deep Networks - dropout, Adam, learning-rate schedules, initialization.
• How LLMs Actually Train - tokenization, cross-entropy, SFT, RLHF, DPO, BPE.
• Scaling & Modern Techniques - scaling laws, batching, distillation, CLIP.
• The Road to Transformers - CNN, RNN, the Transformer arrives.
• Transformer Input Pipeline - embeddings, positional encoding, RoPE.
• Attention - Q/K/V, dot products, classroom and Google-search analogies.
• Computing Attention - scores, K transpose, softmax, scaling, weighted sums.
• Multi-Head Attention - the split, per-head computation, concat, W_O.
• The Encoder - Add & Norm, FFN, residual highway, pre-norm vs post-norm.
• The Decoder - causal masking, cross-attention, output head.
• Modern LLM Techniques - KV cache, grouped-query attention, MoE, reasoning models.
• Vector Search - brute-force kNN, distance metrics, IVF, HNSW, DiskANN.
• Vector Compression - scalar / product / binary quantization, Matryoshka, IVF-PQ.
• Vector DBs in Production - filtering, sharding, replication, hybrid search, rerankers.
• Picking a Vector Database - FAISS, pgvector, Qdrant, Pinecone, Weaviate, Milvus, Chroma.
• The RAG Pipeline - why retrieval, naive RAG, where it breaks.
• RAG Data Prep - parsing, dedup, chunking strategies, contextual retrieval.
• RAG Retrieval - embedding models, hybrid retrieval, rerankers, HyDE, query routing.
• RAG Generation - context packing, lost-in-the-middle, Self-RAG, CRAG, GraphRAG.
• RAG in Production - eval triangle, LLM-as-judge, RAGAS, caching, observability.
• Prompting LLMs - system prompts, few-shot, chain-of-thought, context engineering.
• Tools & Protocols - tool use, MCP architecture, MCP security, A2A protocol.
• Agent Mechanics - agent loop, ReAct, plan-execute, memory taxonomy.
• Multi-Agent Systems - orchestrator-worker, supervisor, hand-offs, critic / debate.
• Shipping Agents - eval, guardrails, prompt-injection defense, LangGraph, CrewAI.

About the author

Rajul Babel built Learn AI as a free, open educational resource. Find more at github.com/rajulbabel and linkedin.com/in/rajulbabel.

Loading interactive app...