AI & Machine Learning Engineering

Welcome to the definitive guide to Artificial Intelligence and Machine Learning Engineering. Rather than treating machine learning frameworks as opaque black boxes, this curriculum derives every core formula from mathematical first principles and implements every algorithm from scratch in both pure Python and pure Rust without third-party dependencies.

Explore curriculum modules from the sidebar to inspect step-by-step mathematical proofs, zero-dependency implementations, interactive zoomable diagrams, and knowledge verification checkpoints.

Curriculum Architecture at a Glance#

The curriculum spans 6 foundational modules covering the full modern AI stack:

1. Supervised Machine Learning Engineering: Derivations and implementations of Logistic Regression, $k$-NN with KD-Trees, Naive Bayes, Support Vector Machines with RBF kernels, CART Decision Trees, AdaBoost and XGBoost gradient boosting, Neural Network matrix backpropagation, and temperature calibration.
2. Unsupervised Learning & Dimensionality Reduction: Lloyd's $k$-Means, DBSCAN density connectivity, Gaussian Mixture Models and EM lower bounds, Principal Component Analysis via SVD, $t$-SNE Student-$t$ KL divergence gradients, and Isolation Forests.
3. Deep Learning, CNNs & Transformer Architectures: Variance-preserving Xavier/He weight initializations, LayerNorm, ResNet gradient highways, LSTM and GRU gating mechanics, Scaled Dot-Product Multi-Head Attention, and LoRA low-rank adaptation.
4. Generative AI, Large Language Models & Diffusion: Variational Autoencoders and ELBO derivations, DDPM forward and reverse diffusion mechanics, Causal Decoder-Only Transformers, KV Cache memory optimizations, Byte-Pair Encoding (BPE), and Direct Preference Optimization (DPO).
5. Reinforcement Learning & Autonomous Agents: Markov Decision Processes, Bellman optimality proofs via Banach contractions, Temporal Difference TD(0) learning, Double DQN decoupled evaluations, PPO clipped surrogate objectives, and the ReAct Autonomous Agent runtime loop.
6. Multimodal AI, Graph Neural Networks & Production MLOps: CLIP InfoNCE dual-encoder contrastive alignment, Graph Convolutional Networks (GCN) with Kipf-Welling renormalization, vLLM PagedAttention virtual memory block allocation, and Post-Training Quantization (PTQ/QAT/AWQ/GPTQ).

Core Pedagogical Standard#

• Step-by-Step Algebraic Derivations: Every loss function, gradient, and optimization update is derived step by step so you understand exactly how equations are formed.
• Zero-Dependency Implementations: Every algorithm is accompanied by complete, standalone implementations in both pure Python (standard library only) and pure Rust (std only).
• Interactive Visual Architecture: Explore interactive flowcharts, mathematical formulas rendered via KaTeX, expandable diagrams, and knowledge verification checkpoints.

The complete course source code and implementations are open on the Source Repository.