MIT OpenCourseWare

MIT OpenCourseWare publishes the teaching materials from more than 2,500 MIT undergraduate and graduate courses, and has done so openly since 2001, making it one of the original large-scale open education projects. A typical course page gathers lecture notes, problem sets and their solutions, exams, reading lists, and often full lecture video with transcripts, arranged the way the course was taught rather than pre-processed for machines.

You reach it through the website, browsing or searching by department and course number, and many courses offer a downloadable archive of their materials. The content is a mix of PDF and HTML, heavy with mathematical notation and diagrams, so getting it into a RAG pipeline means doing your own extraction: pulling text out of PDFs, cleaning it, and splitting it into passages before you embed anything.

For a non-commercial tutoring or study assistant it is hard to beat as a source of rigorous, university-level explanation across science, engineering, economics, and the humanities. The depth and coherence of a full course sequence is exactly the kind of material that makes a teaching RAG system genuinely worth using.

The licence is the part to get right. MIT OpenCourseWare is released under Creative Commons Attribution-NonCommercial-ShareAlike, which imposes three real constraints: you must credit MIT, you cannot use it in a commercial product because the NonCommercial clause rules that out, and any derivative you share must carry the same licence, since ShareAlike propagates. Whether embeddings built from ShareAlike text count as a derivative is genuinely unsettled, so tread carefully before redistributing an index.

It sits alongside OpenLearn and LibreTexts in the open courseware space. If you specifically need material you can build a commercial product on, check those alternatives' licences too, because several carry the same NonCommercial restriction.