# CrackingMachineLearningInterview is a reading path, not a library > CrackingMachineLearningInterview is a Markdown and HTML repository of machine learning interview content organised into role tracks, a question bank, cheat sheets and flashcards. It helps you decide what to study in what order, and it leaves every answer, project and career decision to you. **shafaypro/CrackingMachineLearningInterview** — A repository to prepare you for your machine learning interview, involving most of the questions asked by all the tech giants and local companies. Do this to Ace your Machine Learning Engineer Interviews - Repository: https://github.com/shafaypro/CrackingMachineLearningInterview - Stars: 728 · Forks: 142 - Language: HTML - License: not declared - Published: 2026-10-21 · Updated: 2026-10-21 · Language: en - Canonical page: https://hysenlabs.com/projects/shafaypro-crackingmachinelearninginterview ## The entry point is a question about what you want to build Most interview prep collections are a wall of questions with no order, and the burden of sequencing falls on the reader. This repository attacks that problem first. The README opens with a track chooser at docs/choose-your-track.md, and the instruction is to answer one question about what you want to build, then follow an ordered path through the repository for your role. The tracks named are ML Engineer, AI and GenAI Engineer, Data Scientist, Data Engineer, MLOps and Deep Learning. Each track lists prerequisites, a stage by stage reading order, a project to build, and how to tell when you are interview ready. The audience is broad on purpose, and the README enumerates the roles it serves: Machine Learning Engineer, Data Scientist, Deep Learning Engineer, AI Engineer, a Software Engineer working on AI and ML products, Data Engineer, MLOps Engineer, and DevOps or Platform Engineer. What you get from the repository is navigation and ordering. What you still have to supply is the answers, the implementations and the judgement about whether a given company will actually ask what the track assumes. ## Six topic folders mirror how interviews are actually organised The top level listing shows a repository structured as a curriculum rather than a single document. ai_genai/ holds GenAI and LLM engineering topics including n8n, CrewAI, LangGraph, LangSmith, multi agent systems and advanced RAG. classical_ml/ covers the classical algorithms and the mathematics behind them, linear algebra and optimization, time series, clustering, dimensionality reduction, recommender systems and feature engineering. mlops/ collects MLOps topics, MLflow, model serving, feature stores, explainability, data quality, data labeling and active learning, responsible AI and LLM evaluation. cloud_ml/ maps out AWS SageMaker, Google Vertex AI and Azure ML alongside the core services ML engineers use and a cross cloud service map. data_engineering/ holds data engineering interview topics, platform concepts and geospatial AI. Beyond those the listing shows coding_challenges/, deep_learning/, devops/, frameworks/, project_setup/, system_design/, guides/, a data/ directory, images/, and a Research Papers/ folder. The split tells you something about the author's view of the market: the ML content is the core, and the surrounding folders cover the adjacent roles that hiring managers bundle into the same job description. The primary language of the repository is listed as HTML, which is a small clue about how much of the content is served as pages rather than read as Markdown. ## Two Python tools keep the documentation and the site honest There is code in the repository, and it is aimed at the repository rather than at the reader. tools/check_links.py is an offline link and anchor checker that runs in CI on every pull request. tools/build_site.py generates a standalone, search engine friendly page for every guide in the guides/ directory plus a sitemap.xml. Both are worth knowing about for a practical reason. A repository made of dozens of Markdown files with cross links between them will rot silently as soon as a file is renamed, and the link checker is what turns that into a CI failure instead of a reader complaint. The site generator explains the .nojekyll file, the _config.yml and the sitemap.xml at the top level, all of which are the machinery of a GitHub Pages site rather than of a documentation library. The project also points at its own GitHub Pages site, shafaypro.github.io/CrackingMachineLearningInterview, and describes the interface as having a UI and UX experience, and the flashcards page at flashcards.html is a separate interactive artefact with progress saved in your browser, built from the glossary. There is also practice.html and a MakeMeAwesomeScratch/ directory, and an index.html that presumably serves as the landing page. None of this is documented in the README in detail, so the tools are a place to look if you want to extend the repository rather than merely read it. ## Reading order is prescribed, from roadmap to night before There is nothing to install here, and the README gives no installation steps, because the repository is read in place rather than consumed as a package. What it gives instead is an explicit usage path that is worth following rather than skipping, because the sequencing is the contribution. If you are preparing for current interviews, the 2026 Interview Roadmap at docs/2026-interview-roadmap.md is the starting point. The 2026 Additional Questions and Answers file at docs/2026-additional-questions.md covers modern rounds, with a deeper set in docs/interview_questions_2026.md spanning agents, RAG, LLM scaling, production AI and system design. The Classic Question Bank holds the core content on ML, statistics, deep learning and algorithms, and docs/study-pattern.md sets out recommended preparation topics, difficulty levels and study structure. Preparation Resources and References builds a targeted study plan, and Suggested Learning Order is the clean path from fundamentals to production AI systems. The specialised sections for AI and GenAI, Data Engineering and DevOps cover the narrower tracks. Three files exist for the last stretch. The Behavioral and Project Deep-Dive Guide covers STAR stories, the project deep dive round, ML specific behavioural questions and level expectations, docs/take-home-projects.md explains what reviewers score, how to budget time, how to structure the repository and how the follow up presentation round works, and docs/ml-debugging-scenarios.md is a set of troubleshooting scenarios with ranked causes and fixes for leakage, NaN loss, offline and online gaps, drift and RAG regressions. The README even names the night before: skim the cheat sheet at docs/ml-cheat-sheet.md, drill the glossary flashcards, and rehearse a few debugging scenarios out loud. ## What the repository does not give you The honest limitations are worth stating before anyone plans a study cycle around it. First, this is prose to read, not an assessment. The README describes what reviewers score on take home projects and how the follow up round works, but it does not describe graded exercises with reference solutions inside the repository, so nothing here tells you whether your answer to a given question was right. The 2026 additional questions file is described as a modern question bank covering LLMs, RAG, evaluation, agents and production AI, and the classic bank as the core set, but the README gives no accuracy guarantee for the content and no reviewer attribution for the answers. Second, the licence is not identified, so there is no stated permission to republish the questions, the cheat sheet or the glossary inside your own notes or a company handbook. That is a different constraint from the code, which carries the usual open source treatment, but it applies to the text you would most want to reuse. Third, the 2026 framing dates the content: a bank built around the current interview cycle ages quickly as tooling changes, and the maintainer has no obligation to keep it aligned with what companies ask next year. Fourth, and most practically, the repository cannot judge your readiness. The track says how to tell when you are interview ready, but the signal comes from you, and the same reading order produces a well prepared candidate and an overconfident one. ## How this differs from a course, a textbook or your own question list There are three other ways to prepare, and the difference is about where the effort goes. A paid course gives you a fixed curriculum with videos and exercises, so you pay for structure and get far less freedom to reorder; this repository gives you structure as Markdown, with the freedom to skip a track you have already covered and to spend a week in ai_genai/ or system_design/ without anyone objecting. A textbook gives you depth and rigour in one subject, with the sequencing burden on you and nothing about interviews; docs/ml-cheat-sheet.md and docs/glossary.md are the parts of this repository that come closest to that role, being a one page summary of metrics, losses, distributions and update rules, and every term in the repository defined in a sentence or two with the practical point attached. The third option is to build your own list from the primary sources, the framework documentation and the papers, which is what the Research Papers/ directory in the repository points toward. That route has no curation cost and no staleness, and it takes far longer, because the work the tracks do for you is exactly the ordering and the gap analysis. The practical case for this repository is the middle path: use the track order, then answer the questions yourself and check them against the framework documentation, rather than reading the answers as if they were a rubric. ## Conclusion CrackingMachineLearningInterview suits a candidate who knows roughly nothing about where to start and needs an ordered path, a 2026 question bank and a short list of numbers worth memorising. It does not suit someone looking for graded exercises with reference solutions, since the repository ships prose rather than a checked problem set, and its licence is not identified, so treat the text as something you may read rather than copy into your own notes. Before you plan a study block on it, read docs/choose-your-track.md and one file from the topic folder it points to, then run tools/check_links.py the way CI does to confirm the local files it points at are present in your clone. ## FAQ ### Where should I start in CrackingMachineLearningInterview? The README points new readers at docs/choose-your-track.md, where you answer one question about what you want to build and get an ordered path for your role. From there the 2026 Interview Roadmap covers current AI and ML interview focus areas, and the classic question bank covers core ML, statistics, deep learning and algorithms. ### Does CrackingMachineLearningInterview include practice exercises or solutions? The repository describes take home projects, what reviewers score and how to structure them, and the classic and 2026 question banks, but the README does not describe graded exercises with reference solutions inside the repository. The interactive glossary flashcards are the built-in practice tool. ### What licence is CrackingMachineLearningInterview under? The licence is listed as unknown in the repository metadata and no licence file appears in the top level listing. The project states that you are free to share the repository link in a blog or study notes, but it does not state terms for copying the text itself. ## Sources - [Issues](https://github.com/shafaypro/CrackingMachineLearningInterview/issues) - [README](https://github.com/shafaypro/CrackingMachineLearningInterview/blob/master/README.md) - [shafaypro/CrackingMachineLearningInterview on GitHub](https://github.com/shafaypro/CrackingMachineLearningInterview) --- Hysen Labs editorial analysis, written from the project's own repository and release notes. Cite the canonical page: https://hysenlabs.com/projects/shafaypro-crackingmachinelearninginterview