Let's Learn
A complete reference for Bayesian inference and probabilistic reasoning — the mathematics of updating beliefs with evidence. Bayes’ theorem, priors and posteriors, conditional independence, Bayesian networks, exact vs approximate inference, and how the same 260-year-old theorem powers spam filters, medical diagnosis, robot localization, and modern AI.
A complete reference for the mindset shift every AI practitioner has to make — from notebook thinking (one dataset, one model, one accuracy number) to system thinking (data pipelines, serving infrastructure, monitoring, feedback loops, failure handling, and human oversight). Covers the anatomy of a real AI system, the practical path for making the transition, the core systems-thinking mindsets, real-world examples (recommendations, clinical decision support, fraud, chatbots), and a system-ready checklist for any AI project.
A complete reference for handling imbalanced datasets in machine learning — why accuracy is misleading, which metrics to trust (precision, recall, F1, ROC-AUC, PR-AUC), data-level techniques (SMOTE, random over/undersampling, NearMiss, Tomek links), algorithm-level techniques (class weights, threshold tuning, cost-sensitive learning, focal loss), and how to apply them in fraud, churn, medical, and rare-event problems.
A complete reference for data leakage — the single biggest reason ML models look great in development and fail in production. What leakage actually is, the four common types (target, train-test contamination, temporal, group), how to detect each, and an audit checklist that prevents it across the entire pipeline.
A complete reference for cross-validation strategies beyond plain K-Fold — hold-out, stratified K-Fold, leave-one-out, time-series walk-forward, group K-Fold, nested CV, and repeated Monte-Carlo CV. How to choose the right scheme for your data, the common mistakes that inflate scores, and a practical decision flow for real-world ML projects.
A 10-chapter practitioner-oriented guide for software engineers and architects building AI agents in production — agent architecture and design patterns, tool use and function calling, memory and state management, planning and reasoning loops, multi-agent orchestration, evaluation and safety, observability and ops, plus real-world reference implementations from Meta, Amazon, and GitHub.
A complete reference for the bias-variance tradeoff — what bias and variance actually are, why underfitting and overfitting are two faces of the same coin, the math behind the decomposition, how to detect each failure mode using learning curves, and the practical fixes that work in real-world ML pipelines.
A 10-chapter engineering-first course on building AI agents that survive production — agent architecture patterns, tool use and function calling, memory and state, planning and reasoning loops, multi-agent orchestration, evaluation, safety, scaling, and the observability and operational practices that separate a working demo from a system you can actually operate.
A 13-chapter advanced AI course for practitioners — advanced deep learning architectures, large-scale training and distributed systems, RLHF and alignment, retrieval-augmented generation, agentic systems, multimodal models, model efficiency, evaluation and AI safety, and the open research frontiers shaping the next generation of systems.
A practitioner-oriented 15-chapter intermediate AI course covering applied foundations, advanced ML, sequence models and attention, LLM internals, prompting and RAG, generative media, autonomous agents, MLOps, computer vision in production, end-to-end system design, and AI safety and research frontiers.