DSprep.com

About

DSprep is a free, open-source study base for data science interviews, maintained by SourceStrongAI. It collects interview questions and answers across 17 topics — from data structures, machine learning, SQL, probability, and A/B testing through the newer rounds: Transformers, LangChain, LangGraph, Forward Deployed Engineer, and generative-AI system design. Around the banks sit 14 cheat sheets, 20 ML algorithms explained and implemented from scratch, a flashcard drill, and seven full courses. All of it is readable on a first visit: the site runs without accounts and never asks for an email address.

It meets candidates at every stage, and says so structurally: four ordered learning paths — Foundations First for newcomers, then Data Scientist, then Machine Learning Engineer, then GenAI/LLM Engineer — arrange the material into routes of roughly four to ten weeks, each closing with a checklist of what you should be able to do before moving on. Content is organized by topic, not by company.

The plumbing is deliberately plain: pages are written in Markdown, compiled to a static site with MkDocs and the Material theme, hosted on GitHub Pages, and searchable without a server round-trip. Maintenance is communal under an AGPL-3.0 license, and every page carries an edit link. New material arrives from readers' own interviews: walk out of one, write down what you were asked, and send the raw file — .txt, .md, .py, or .ipynb all work, because turning rough notes into a finished page is the maintainers' job, not yours. Curation runs the other way too — questions get pruned once they stop showing up in real interviews.

The seven courses go deepest: one per design round — modern systems, ML systems, generative AI, mobile, frontend, product architecture — plus a 48-hour fast-track option, with estimates of 18 to 34 hours per course. Each follows the same published 18-section structure: prerequisites you can test yourself against, time budgets with the arithmetic shown, four-criterion rubrics, scripts a non-expert friend can run a mock round from, timed drills, a catalog of common mistakes, and a dated changelog. The site is also explicit about a limit: a page of questions cannot do what a human interviewer does in a mock round. It stays free on principle; if you feel like paying anyway, the FAQ suggests giving that money to a charity.

Highlights

• Q&A banks across 17 topics, including Transformers, LangChain, LangGraph, Forward Deployed Engineer, and generative-AI system-design rounds
• Seven free system-design courses, one per interview round, estimated 18-34 hours each — with rubrics, interviewer scripts, timed drills, and changelogs
• Four role-based learning paths (4-10 weeks each) that end in concrete readiness checklists
• 14 cheat sheets, 20 ML algorithms implemented and documented from scratch, and a flashcard drill tool
• Open source under AGPL-3.0: every page has an edit link, and raw contributions land through GitHub as .txt, .md, .py, or .ipynb files

Who it’s for

For anyone working toward data science and adjacent interviews: beginners who need an ordered route in, working data scientists sharpening up for one specific round, and engineers targeting ML or GenAI/LLM roles. It also serves the other side of the loop — people fresh out of an interview who want to contribute the questions they faced, since the whole site is maintained through GitHub by its readers.

Details