AI / ML / MLOps Interview Handbook
A complete interview-prep handbook for ML Engineer, MLOps Engineer, and AI Engineer roles — rebuilt to be understandable, not just dense. It is ordered the way the field was actually invented: we start with the mathematics that everything rests on, walk forward through classical machine learning and deep learning, reach the transformer and the modern LLM, and end at today’s frontier — RAG, agents, and the loops that wrap models into products. A final practical-toolkit set covers the libraries you get grilled on by name, closing with deploying to the cloud — AWS Bedrock, Google Vertex AI, and AI Studio.
Every topic follows the same shape: a short plain-language explainer (intuition first), then simple question-and-answer drills with simplified answers, and a diagram, formula, snippet of code, or small visual wherever it makes the idea click.
How to use this: each chapter is its own page (so it loads fast). Pick one below, read the explainers to build the model, then cover the answers and quiz yourself on the Q: lines. Every chapter page has prev / next links and a jump-to-any-chapter menu at the top.
🗺️ Chapters — the chronological path
🧮 Mathematical Foundations — the bedrock every model stands on
- 01. 🧮 Linear Algebra — the language of data
- 02. 📉 Calculus & Optimization — how models learn
- 03. 🎲 Probability & Statistics — reasoning under uncertainty
- 04. 🔥 Information Theory & Loss Functions — measuring surprise and error
🧩 Classical Machine Learning — the algorithms that ran ML before deep learning
- 05. 🧩 Core ML Concepts — the ground rules
- 06. 📐 Classical Supervised Algorithms — the workhorses
- 07. 🌲 Ensembles & Boosting — how to win on tabular data
- 08. 🗺️ Unsupervised Learning & Dimensionality Reduction — structure without labels
- 09. 🎯 Model Evaluation & Validation — knowing if it actually works
🧠 Deep Learning — neural networks, and how to actually train them
- 10. 🧠 Neural Network Fundamentals — the building block
- 11. ⚙️ Training Deep Networks — making deep nets actually train
- 12. 🖼️ Convolutional Neural Networks — the vision branch
- 13. 🔁 Sequence Models — RNNs, LSTMs and the bottleneck
⚡ The Transformer Era — embeddings, attention, and the architecture behind every LLM
- 14. 🔤 Word Embeddings — giving words meaning as vectors
- 15. ⚡ Attention & the Transformer — the architecture that changed everything
- 16. 🧱 Tokenization, Pretraining & Model Families
- 17. 📈 Modern LLMs & Scaling — bigger, and suddenly capable
💬 Using & Adapting LLMs — prompting, fine-tuning, RAG, and serving
- 18. 💬 Prompting & In-Context Learning — programming models with words
- 19. 🎚️ Fine-Tuning & Alignment — specializing and aligning models
- 20. 📚 Retrieval-Augmented Generation (RAG) — giving the model an open book
- 21. 🚀 Inference, Decoding & Serving — running LLMs efficiently
🤖 The Agentic Frontier — agents, safety, and operating it all in production
- 22. 🤖 Agents, Tools & Loops — the latest frontier
- 23. 🛡️ Evaluation, Safety & Guardrails — making LLM systems trustworthy
- 24. 🔧 MLOps & LLMOps — shipping and operating models in production
🛠️ The Practical Toolkit — the libraries and tools you will be asked about by name
- 25. 🛠️ Practical Toolkit I — Modeling & Vision Libraries
- 26. 🧰 Practical Toolkit II — LLM Frameworks, Orchestration & Vector Stores
- 27. ⚙️ Practical Toolkit III — Serving, Apps & MLOps Tooling
☁️ Cloud AI Platforms — deploying foundation models on AWS, Google, and Azure