3 free AI engineering tracks Stanford says you should be doing

What you'll walk away with: Three free self-paced tracks that actually move your AI engineering skills (not just "awareness"), an 8-week sequence to do them in, a Stanford-citable CV bullet, and one active paid opportunity with a hard deadline.

Prerequisites: Laptop, internet, ability to follow instructions. No maths PhD, no prior AI work.

Time to complete: Reading this — 8 minutes. The 8-week track — 4-6 hours per week.

What Stanford actually said

The Stanford HAI 2026 AI Index report, published in April 2026, looked at which countries' workers are learning AI engineering skills fastest based on platform data from Coursera, DeepLearning.AI, and similar providers.

The verbatim finding:

"The United Arab Emirates, Chile, and South Africa are learning AI engineering skills fastest."

Not India. Not China. Not the US. The three countries with the steepest AI engineering skill acquisition rate globally were UAE, Chile, and South Africa.

Source: Stanford HAI AI Index 2026

What this doesn't mean: SA is ahead on AI overall. The US still leads on model releases, compute, and research output.

What it does mean: the skills pipeline in SA is moving faster than in most of the world. If you're actively upskilling, you're part of a statistical cohort Stanford has now named. That's a reference you can use.

Why this matters for your CV and job hunt

Two uses, both immediate:

1. CV / LinkedIn credibility anchor. "AI engineering skills" on a CV usually reads as self-promotion. Cited alongside a Stanford report, it reads differently. Recruiters scanning quickly recognise Stanford HAI as a legitimate source.

2. Identity signal in interviews. "I'm part of one of the three fastest-learning AI engineering cohorts globally according to Stanford's 2026 AI Index" is a stronger opener than "I'm interested in AI."

Neither works if you haven't actually done the learning. This guide is the "actually do it" part.

The 8-week sequence

graph LR
    A["Week 1-2"] --> B["Track 1:<br/>DeepLearning.AI"]
    C["Week 3-5"] --> D["Track 2:<br/>Anthropic Learn"]
    E["Week 6-8"] --> F["Track 3:<br/>fast.ai"]
    B --> G["Foundation"]
    D --> H["Application"]
    F --> I["Build + ship"]
    G --> J["CV bullet<br/>unlocked"]
    H --> J
    I --> J
    style A fill:#18181B,stroke:#FFAA00,color:#F5F5F7
    style C fill:#18181B,stroke:#FFAA00,color:#F5F5F7
    style E fill:#18181B,stroke:#FFAA00,color:#F5F5F7
    style J fill:#18181B,stroke:#FFAA00,color:#F5F5F7

Track 1 — DeepLearning.AI Short Courses (foundation)

Link: deeplearning.ai/short-courses Cost: Free Time: 1-2 hours per course, 3 courses to start When: Weeks 1-2

DeepLearning.AI (founded by Andrew Ng) runs short free courses in partnership with the labs that build the tools. Every course has a hands-on notebook.

The three to do in this order:

1. ChatGPT Prompt Engineering for Developers — 1.5 hours. Co-taught with OpenAI. Teaches the actual prompt patterns (chain-of-thought, few-shot, delimiters, output structuring).
2. Building Systems with the ChatGPT API — 1.5 hours. How to build multi-step AI applications. Moves you from "I use ChatGPT" to "I build with the API."
3. LangChain for LLM Application Development — 1.5 hours. The framework most production AI apps are built on. LangChain shows up on job descriptions constantly.

Total: about 4.5 hours. You now understand the mental model of how AI applications are constructed, not just used.

Tip: Don't just watch — open the Jupyter notebook in a second tab and run every cell. Passive watching leaves nothing on your CV.

Track 2 — Anthropic Learn (application)

Link: anthropic.com/learn (and search for "Anthropic Academy" / the Claude Skills course) Cost: Free Time: 3-5 hours per track When: Weeks 3-5

Anthropic (the company behind Claude) runs free technical courses specifically around building with Claude. These are closer to what AI engineers get paid to do than generic "intro to LLMs" content.

The three that matter for jobs:

1. Prompt Engineering Interactive Tutorial — GitHub-based. Clone the repo, work through the notebook exercises. Tests your ability to write actually functional prompts. Takes a solid afternoon.
2. Building with Claude — tool use and agents — the agent pattern is what every AI company wants you to know right now. Covers MCP (Model Context Protocol), tool calling, and multi-turn planning.
3. Computer Use / Claude Code — newer track. How to make Claude operate a computer or a codebase on its own. This is literally the 2026 job description for a junior AI engineer.

By the end of Track 2, you've built small working things. You've committed code. You have real examples to reference in interviews.

Track 3 — fast.ai (build and ship)

Link: course.fast.ai Cost: Free Time: 20-30 hours for the core course When: Weeks 6-8

fast.ai is the builder-focused track. Less lecture, more shipping. Jeremy Howard's approach is famously "build a working thing on day one, understand the maths later."

Don't try to do all of it. Prioritise:

1. Lessons 1-2: Image classifier from scratch. You ship a working classifier by end of lesson 1.
2. Lesson 3: Production deployment. Your classifier runs as a web app. This is the portfolio piece.
3. Lessons 4-5: NLP. Text classification and sentiment analysis with working code.

Skip the deeper maths lessons (7-8) if you're tight on time. You can come back later. By end of Track 3 you have a GitHub repo with a deployed AI app. That's the thing recruiters actually look at.

Done when: you have one deployed AI app, one GitHub commit history showing consistent work, and one written post (LinkedIn, Medium, or a blog) explaining what you built.

The Stanford-citable CV bullet (exact wording)

Use one of these two on your CV or LinkedIn. Both cite Stanford properly and both read as credible.

Variant 1 — skills-forward:

Building AI engineering skills via DeepLearning.AI, Anthropic, and fast.ai tracks — part of the cohort Stanford HAI's 2026 AI Index named as globally top-3 for fastest AI engineering skill acquisition (UAE, Chile, South Africa).

Variant 2 — outcome-forward (use after Track 3):

Shipped [your app name] — [one-line description of what it does]. Built during self-directed AI engineering learning across DeepLearning.AI, Anthropic Academy, and fast.ai. Part of the Stanford HAI 2026 AI Index top-3 cohort for fastest AI engineering skill acquisition globally.

Do NOT write "Stanford-ranked AI engineer" or similar overclaims. Stanford ranked the COUNTRY, not you individually. Phrase it as cohort membership, which is accurate.

LinkedIn headline rewrite

Replace this:

"Aspiring AI engineer | Student at UCT"

With this:

"Building AI engineering skills | Stanford AI Index 2026 named SA top-3 globally for fastest AI skill acquisition | Shipping at [github-link]"

It reads stronger because (1) it cites a real source, (2) the signal is "actively building" not "aspiring", (3) the GitHub link is proof.

The one opportunity with a hard deadline — Anthropic Fellows

Applications close: April 26, 2026.

Anthropic runs a Fellows programme that pays USD $3,850 a week for AI alignment and engineering research. That's approximately R70,000 per week at current rates, or R285,000 a month. South Africans are eligible. You don't need a PhD — you need demonstrated engineering chops and genuine interest in safety research.

If you're on Track 1-2 of this sequence and the deadline is this month, applying now with an "in progress" framing is legitimate. You show the work you've shipped so far, explain the trajectory, and apply. Your baseline outcome is a rejection you learn from. Your upside is R285K/month for six months.

Application link: Search "Anthropic Fellows 2026" on Anthropic's careers page. Alternative: globalsouthopportunities.com coverage.

When you have 2 weeks, not 8

Minimum viable version:

Time available	Do this
2 weeks	Track 1 only (4.5 hours) + deploy one of the notebooks publicly. Use Variant 1 CV bullet.
4 weeks	Tracks 1 + 2. Ship one demo. Use Variant 1 bullet.
8 weeks	All three tracks. Ship a portfolio app. Use Variant 2 bullet. Apply to Anthropic Fellows if deadline hasn't passed.

Any of these is better than zero. The point is to not be in the "aspiring" group anymore.

What this guide doesn't do

• Get you a job by itself. Shipping a portfolio app + applying to 30+ roles + interview prep is still required.
• Replace a computer science degree. It complements one.
• Work if you don't actually do the courses. The Stanford stat is about learners, not about people who bookmarked the link.

Everything above takes 40-60 hours of focused work. That's 6-8 weeks at 5-8 hours per week. You can watch 40 hours of TikTok doing it.

Recap

• Stanford HAI 2026 AI Index named UAE, Chile, and South Africa top-3 for fastest AI engineering skill acquisition globally
• This is a citable credibility anchor you can use on CV and LinkedIn — but only if you do the learning
• 8-week sequence: DeepLearning.AI (foundation), Anthropic Learn (application), fast.ai (build and ship)
• Stanford-citable CV bullets use cohort framing, not individual ranking
• Anthropic Fellows programme pays $3,850/week, SA eligible, closes April 26 — apply with in-progress framing
• Ship one deployed app. GitHub link beats every CV claim.

This guide powers the "Comment STANFORD" CTA — you got it because you asked for it.