Google DeepMind's main structured course offering is AI Research Foundations, a free eight-course curriculum on Google Skills that teaches how language models are built, from tokenization to fine-tuning. It was developed with University College London and is aimed at university students and community learners who already know Python and study computer science, math, physics, or a related subject. DeepMind also runs teacher resources and funded scholarships, but those are separate from this curriculum.
Key takeaways
- AI Research Foundations has eight courses, and the listed hours for courses one to seven add up to 38, plus 20 or more for the capstone.
- The courses are described by Google as available at no cost on Google Skills, and the lab code is public on GitHub under the Apache 2.0 license.
- Python and a technical background are the stated entry point, not a DeepMind-specific qualification.
- Experience AI, a separate program with the Raspberry Pi Foundation, serves secondary school teachers, not university learners.
- Master's scholarships and postdoctoral fellowships exist, but they are funded places, not online courses.
What AI Research Foundations covers
The curriculum is hosted as a learning path on Google Skills, which describes it as covering the technologies behind language models such as Gemini and how to build and fine-tune them from the ground up. UCL reported its launch on October 21, 2025. Google's blog post of March 17, 2026 adds that the courses are free and that UCL pedagogy experts and academics helped develop them. Each course mixes reading, videos, labs, and short assessments, and the lab code sits in the public google-deepmind/ai-foundations repository, with software under Apache 2.0 and other materials under CC-BY 4.0.
The repository carries a warning worth repeating: the language models built in the labs are for educational use, can produce inaccurate or inappropriate output, and should not be relied on for medical, legal, or financial advice.
The eight courses and their hours
The African Institute for Mathematical Sciences (AIMS), which runs a lecturer-training program around the curriculum, publishes the following titles and time estimates. The hours are AIMS estimates and may differ from what a learner needs.
| Course | Title | Estimated hours |
|---|---|---|
| 1 | Build Your Own Small Language Model | 6 |
| 2 | Represent Your Language Data | 4 |
| 3 | Design and Train Neural Networks | 4 |
| 4 | Discover the Transformer Architecture | 4 |
| 5 | Fine-tune Your Model | 8 |
| 6 | Align Your Model | 5 |
| 7 | Accelerate Your Model | 7 |
| 8 | Capstone Project | 20 or more |
A learner who wants only the core mechanics could stop after the first four courses, which total 18 hours. That figure matches the pre-work AIMS asks of lecturers before its workshop. The path page on Google Skills lists the eight items but not durations, prices, or lab credit rules, so anyone planning to enroll should read the current terms on that page before starting.
Other DeepMind learning programs
On its education page, Google DeepMind lists programs that are not online courses:
- Experience AI, launched in April 2023 with the Raspberry Pi Foundation, offers lesson plans, slide decks, and videos for secondary school teachers; the page says it has been accessed in 180 countries and is available in 19 languages.
- AI Masters scholarships are delivered in the United Kingdom through the Martingale Foundation in 2026, with applications closed at the time of checking, and through the Institute of International Education for host universities elsewhere.
- A Pan-African AI for Science Master's, created with AIMS and offered at AIMS South Africa, aims to support 160 African students by July 2027.
- Academic chairs in machine learning exist at the University of Alberta, the University of Cambridge, the University of Oxford, and UCL.
The Alberta chair is the main Canadian link on that page. Experience AI and the scholarships are not designed as self-paced courses for adult learners in Canada or the United States.
Free recorded lectures
Before the curriculum existed, DeepMind and UCL published a lecture series on YouTube titled "DeepMind x UCL | Deep Learning Lecture Series 2020." It is older material and predates large language models in their current form, so it suits learners who want foundations in deep learning rather than a language-model build path.
The Africa program and what it signals
Google's philanthropic arm provided USD 4 million to the FATE Foundation to fund lecturer training and teaching toolkits, according to Google's March 17, 2026 post. AIMS lists a goal of reaching 30 universities through 2028: 10 in Nigeria, 10 in South Africa, six in Kenya, and four in Ghana. For learners elsewhere, the practical point is that the content is designed to be taught in classrooms as well as taken alone, and that UCL and AIMS built a companion toolkit for universities that want to teach from it.
FAQ
Are DeepMind courses free?
Google describes the AI Research Foundations courses as available at no cost on Google Skills. Lab access rules can change, so the current terms on the path page decide.
Do the courses lead to a DeepMind degree?
No source reviewed describes a degree. The path is a set of online courses, and any certificate or badge terms appear on Google Skills.
What background is needed?
The stated audience is university students and community learners who know Python and have a background in computer science, math, physics, or another technical subject. AIMS adds that linear algebra, calculus, and probability help but are not required.
Where can teachers find DeepMind material?
Experience AI, run with the Raspberry Pi Foundation, provides lesson materials for secondary school teachers. It is listed on the DeepMind education page.
Related reading
- AlphaGo and the game of Go
- DeepMind papers at ICLR 2023
- How much math AI work needs
- Apps for learning AI skills
More in DeepMind.
Last verified: October 2026
Sources
- Google DeepMind: AI Research Foundations, Google Skills, 2026
- UCL joins forces with Google DeepMind to democratise access to AI education, University College London, October 21, 2025
- Broadening advanced AI education across Africa, Google, March 17, 2026
- AI Research Foundations for Higher Education program page, African Institute for Mathematical Sciences, 2026
- google-deepmind/ai-foundations repository, GitHub, 2026
- Education, Google DeepMind, 2026
- DeepMind x UCL Deep Learning Lecture Series 2020, YouTube playlist, 2020