Most “best AI certifications” lists give you a menu and walk away. That’s the problem. AI certifications work like a tech stack — the order matters more than the list. Take them in the wrong sequence and you’ll repeat material, burn weekends, and still not have anything coherent on your resume. Take them in the right sequence and each one teaches you something the next one assumes you know.
This guide is the 90-day stack: five legitimate, free (or audit-free) AI certifications that build on each other, with specific weeks for each, what to skip, and how to put them on your LinkedIn so a recruiter actually clicks.
Quick disclosure on “free”: “Free to audit” means you can take the course materials at no cost on Coursera, edX, or YouTube. Some certs cost a small exam fee ($0–$99) for the official paper certificate. Where that applies, we’ll flag it. Three of the five below are 100% free.
The 90-day stack at a glance
| Week | Certification | Cost | Hours | What it adds |
|---|---|---|---|---|
| 1–3 | Hugging Face NLP Course | Free | 30 | Foundational understanding of transformers, tokenizers, fine-tuning |
| 4 | DeepLearning.AI — Generative AI with LLMs | Free (audit) | 16 | Andrew Ng’s framework for thinking about LLMs in production |
| 5–6 | Google Cloud Generative AI Leader | $99 exam | 25 | First vendor cert; cloud strategy for AI |
| 7–9 | AWS AI Practitioner OR Microsoft AI-900 | $99–$100 exam | 40 | The cert recruiters actually search for |
| 10–13 | IBM AI Engineering (Coursera) | Free (audit) or $59/mo | 60 | Capstone with portfolio projects |
Total time: ~170 study hours, spread across 13 weeks. That’s about 13 hours per week, or 1.5 hours daily. Doable around a job or full-time school.
Week 1–3: Hugging Face NLP Course (foundations)
Start here, not with a vendor cert. The Hugging Face NLP Course is the only free resource that teaches you what’s actually under the hood of every commercial AI tool — tokenizers, transformer architecture, attention, fine-tuning. Every later course will assume this knowledge. Skip it and you’ll memorize “RAG” and “embeddings” without understanding why they work.
What to actually do in 3 weeks:
- Week 1: Chapters 1–4 (transformer models, using pre-trained models). Run every code example in Google Colab.
- Week 2: Chapters 5–7 (datasets, tokenizers, fine-tuning). Fine-tune one model on a Kaggle dataset of your choice — this becomes your first portfolio piece.
- Week 3: Chapters 8–9 (debugging, sharing models). Push your fine-tuned model to the Hugging Face Hub. Free. Public. Linkable.
What to skip: The audio and computer vision chapters (10–11) unless those are your specialty. They’re great but slow you down for the stack.
The certificate: There isn’t one in the traditional sense. Hugging Face gives you a profile with the models you’ve published and the courses you’ve completed. That’s actually better than a PDF certificate — it’s a live portfolio link that recruiters can click.
Week 4: DeepLearning.AI — Generative AI with LLMs (the framework)
Andrew Ng’s short course on Coursera (16 hours, free to audit, ~$49 if you want the certificate). After Hugging Face, you understand the parts. This course gives you the framework for thinking about LLMs as products: scaling laws, RLHF, prompt engineering, fine-tuning vs RAG decisions.
It’s the bridge between “I can run a model” (HF) and “I understand why this model exists in this form” (vendor certs).
One week is enough. The course is split into 3 weeks officially but you can compress it into one if you watch at 1.5x speed and skip the optional readings. Audit mode if you don’t need the paper.
Week 5–6: Google Cloud Generative AI Leader (your first paper cert)
Now you take a vendor exam. Google’s GenAI Leader cert is the easiest of the three major-cloud entry-level AI certs. It’s strategy-focused — less technical than AWS or Azure equivalents — which means after Hugging Face + DeepLearning.AI you’ll find it almost too easy.
That’s the point. You want a quick “first vendor win” that gives you confidence and a real certificate to put on LinkedIn before you tackle the harder ones.
- Cost: $99 exam fee.
- Study materials: Google’s free Skills Boost path (about 20 hours).
- Format: 50 multiple-choice questions, 90 minutes, online proctored.
- Pass rate: Around 75–80% for prepared candidates.
Week 7–9: AWS AI Practitioner OR Microsoft AI-900 (the recruiter-search cert)
Pick one based on which cloud your target employers use. Both certs are roughly equivalent in difficulty and value, but recruiters search for them by exact name in LinkedIn searches.
- Pick AWS AI Practitioner if you’re targeting US tech companies, startups, or any company already on AWS. Full study guide here.
- Pick Microsoft AI-900 if you’re targeting enterprise jobs, healthcare, finance, or government — sectors that lean Microsoft. Detailed comparison here.
- Don’t pick both. They overlap ~70% in content. Doing both is a waste of two weeks for resume value of one.
If you’re genuinely undecided, use our 60-second AI Certification Recommender — it weighs your goals, budget, and cloud preference and tells you which one fits.
Week 10–13: IBM AI Engineering Professional Certificate (the capstone)
You finish with a heavyweight: IBM’s 6-course Coursera specialization. About 60 hours of work spread across 4 weeks. You can audit it for free or pay $59/month for the certificate (one month is usually enough if you push).
Why this last and not first? Because it has a capstone project. You’ll build a working ML/AI application end-to-end. After 9 weeks of foundations and vendor certs, you have the conceptual base to actually finish the project well — not just submit a Jupyter notebook copy-paste.
The capstone is your second portfolio piece (after the Hugging Face fine-tune). Two real projects on a public profile beat any number of certificates with no code attached.
How to put the stack on your resume
Don’t list 5 certifications as 5 lines. Group them with a project context. Example:
AI / Machine Learning Foundation (Q1–Q2 2026)
Completed structured 90-day learning path covering NLP fundamentals (Hugging Face), generative AI (DeepLearning.AI), cloud AI strategy (Google Cloud Generative AI Leader), production AI services (AWS AI Practitioner), and end-to-end ML engineering (IBM AI Engineering, Coursera).
Portfolio: [link to Hugging Face profile] · [link to capstone project repo]
This frames the stack as a deliberate learning program, not a Coursera shopping spree. Recruiters scanning for “AWS AI Practitioner” still find the keyword. Hiring managers reading the bullet see a coherent journey.
Common mistakes that waste 30+ hours
- Starting with a vendor cert. You’ll memorize service names without understanding what they do. The Hugging Face foundation makes the vendor cert click instead of feel like flashcard memorization.
- Doing both AWS and Azure entry certs. They cover the same concepts with different brand names. Pick one.
- Paying for the DeepLearning.AI certificate. The course is the value; the certificate is just a $49 PDF. Audit it.
- Skipping the Hugging Face fine-tune project. “Took the course” is not a credential. “Published a fine-tuned BERT for sentiment analysis on Hugging Face Hub” is.
- Doing them in random order. The IBM specialization assumes you understand transformers. Start there and you’ll spend half your time looking up basics.
What if you only have 30 days?
Compressed plan: Hugging Face chapters 1–4 (week 1) → Google Cloud GenAI Leader (week 2) → AWS AI Practitioner (weeks 3–4). Skip the IBM capstone for later. You’ll have 3 certs and a basic foundation, but no portfolio project. Better than nothing if you have a deadline.
Frequently asked questions
Are these 5 certs really enough to get hired?
For entry-level AI/ML roles or for adding AI skills to a tech-adjacent role (e.g., backend dev, data analyst), yes — combined with the two portfolio projects from the stack. For senior or specialist AI engineering roles, you’d need an engineer-level cert (AWS MLA-C01 or Azure AI-102) plus original projects beyond the courseware.
Can I do this stack while working full-time?
Yes. 13 hours a week is roughly 2 hours weekday evenings + half a Saturday. Most readers report finishing in 14–16 weeks instead of 13 when balancing a job, which is fine.
What if I’m in Pakistan / India / Africa — are these certs recognized?
The four English-language certifications (Hugging Face, DeepLearning.AI, Google, AWS, Microsoft, IBM) are globally recognized by remote-first companies. The exam fees are USD-denominated — Google offers regional discounts in some countries via Skills Boost partnerships. More on remote tech jobs from Pakistan in 2026 here.
Should I add any specialized certs (computer vision, robotics)?
Only after this base stack and only if your target job requires them. Specialization without foundation is a common trap — you’ll know “YOLO object detection” but not why CNNs work. Get the base in place first.
Start the stack today
Open the Hugging Face NLP Course in one tab and Google Colab in another. Read chapter 1 today. Run the first code example. You’re now 1 hour into a 170-hour stack — the only one that matters is the one you’ve started.
If you want a custom recommendation tailored to your goal, budget, and timeline rather than this general stack, take the 60-second AI Certification Recommender — it’ll point you to the best starting cert based on your specific situation.
