The AI Career Playbook

01How to use this

This is a 30-day plan. You build three projects, each one harder than the last, each one live on a real URL when you finish it. You need a laptop, an internet connection, and roughly two focused hours a day. Every tool in this playbook has a free tier that covers everything you'll do here.

Read the whole thing once (about 20 minutes), then come back and work through it day by day. Don't read it five times instead of starting. Day 1 is opening a GitHub account, not mastering anything.

One honest warning: 30 days of this is uncomfortable. You will be confused on day 3, annoyed on day 12, and stuck on day 22. That's the plan working. The people who get hired are the ones who kept going through exactly that.

02The rule the whole plan rests on

Proof beats certificates. A recruiter scanning 200 resumes stops at the one with a link they can click and a thing that actually works. Certificates say you attended. A live URL says you can build.

So every project in this plan ends the same way: deployed, public, with a link. Not "done on my laptop." Not "90% finished." Live, or it doesn't count. That single standard is what separates this from watching tutorials.

03The 30 days at a glance

Three projects is deliberate. One project looks like a fluke. Three, in rising difficulty, is a track record — and it gives you three different stories to tell in an interview.

04Days 1–7 · Project 1: a portfolio site on a real URL

Your portfolio site is the container for everything else. Projects 2 and 3 will live on it. It's also the gentlest way to learn the full loop: write code → push to GitHub → see it live.

The day-by-day

• Day 1: Create a free account on GitHub. Install VS Code. That's it. Day one is setup.
• Day 2: Open a free AI assistant — Claude or Gemini — and ask it to explain what an HTML file is, then to generate a simple one-page site about you. Read the code it gives you. Ask it what every part does. This "generate, then interrogate" habit is how you learn fast with AI instead of learning nothing with AI.
• Day 3–4: Build your page for real: your name, one line about what you're becoming, a projects section (empty for now — it fills up over the next 23 days), and how to reach you. Change the AI's design until it feels like yours. When you don't understand a CSS property, look it up on MDN — it's the reference professionals use.
• Day 5: Put the code on GitHub. Ask your AI assistant to walk you through git init, commit, and push — one command at a time.
• Day 6: Deploy. Two free routes: GitHub Pages (simplest for a static page) or Vercel (what you'll use for project 2 anyway). Follow their setup guide; it's under 30 minutes either way.
• Day 7: Polish on a phone. Open your live URL on your own phone and fix what looks broken. More people will see your site on a phone than a laptop.

05Days 8–19 · Project 2: a real tool with login and a database

This is the project that changes how interviewers see you, because it has the three things real software has: users, data, and state. Pick a problem from your own life — you'll make far better decisions about a tool you'd actually use.

Three ideas that work (pick one, or adapt)

• Expense splitter for roommates. Log shared expenses, see who owes whom, settle up. Features: sign in, add expense, running balance per person.
• Appointment book for a local business. A salon, a tuition teacher, a physio you know. Customers pick a slot; the owner sees the day's list. Bonus: you now have a real user to interview.
• Study tracker with streaks. Log what you studied, see your streak, and a weekly summary. Simple data model, satisfying to use daily.

The stack (all free)

Frontend: HTML/CSS/JavaScript (or React if the AI-generated code pulls you there — either is fine at this stage). Database and login: Supabase — its free tier gives you a real Postgres database and ready-made authentication. Hosting: Vercel, same as project 1.

How to spend the 12 days

• Days 8–9: Sketch the screens on paper or Excalidraw. List the data you're storing. Write both down before touching code — a day of thinking saves three days of rework.
• Days 10–13: Build the core screen without login: adding and listing your data, saved in Supabase. Get one full loop working end to end before anything else.
• Days 14–16: Add sign-up and sign-in with Supabase Auth. Make each user see only their own data. This step is where most tutorials stop — which is exactly why interviewers probe for it.
• Days 17–18: Deploy to Vercel. Fix what breaks in production; something always does, and fixing it teaches you more than the build did.
• Day 19: Add the link and a screenshot to your portfolio site. Write a README that says what it is, what it's built with, and what you'd add next.

06Days 20–30 · Project 3: an AI agent that runs on its own

An agent is a program that does a job without you clicking anything: it takes an input, runs AI reasoning on it, and produces a useful output. In 2026 this is the most asked-about skill in interviews, and almost nobody entry-level can show one working. You will be able to.

Three ideas that work

• Resume vs job-description checker. Paste a JD and a resume; the agent scores the match, lists missing keywords, and suggests two bullet rewrites. (You'll also use this yourself in week 5.)
• Document Q&A bot. Feed it one document — a syllabus, a product manual, your college's fee rules — and it answers questions from that document only, saying "not in the document" when it isn't.
• Daily digest agent. Every morning it fetches a few sources on a topic you care about, summarizes them into five bullets, and emails you. Scheduled, automatic, no button.

The stack (all free)

The Gemini API has a free tier generous enough for all three ideas; Google's AI Studio gives you a key in minutes. Write the agent in Python or JavaScript — whichever you've been leaning on. Host it on Vercel (as an API route) for the first two ideas; for the scheduled digest, a GitHub Action on a timer is the free, professional way.

How to spend the 11 days

• Days 20–21: Get one API call working: send text to the model, print the reply. Small, but this is the moment you stop being an AI user and become an AI builder.
• Days 22–25: Build the core logic. Spend real time on the prompt — write down five test inputs and what a good output looks like for each, then tune until it passes all five.
• Days 26–28: Wrap it: a simple page for idea 1 or 2, a schedule for idea 3. Handle the ugly cases — empty input, a huge document, the API timing out.
• Days 29–30: Deploy, test with someone else's input (not yours), then add it to your portfolio with a 60-second demo description: the problem, what the agent does, one example input and output.

07The free toolbox

That's the whole list. If a course or tool asks for money in your first 30 days, you don't need it yet.

08Putting the projects on your resume

Projects go in their own section, above education if you're a fresher — they're your strongest evidence, so they go first. Each project gets two to three lines using this shape:

Compare:

Where do numbers come from at your stage? Count what's true: users ("my 4 roommates"), scale ("handles 200+ logged expenses"), speed ("shipped in 12 days"), scope ("3 deployed projects in 30 days"). Small honest numbers beat big vague claims — interviewers can smell inflated ones.

Say the AI part plainly: "built with AI-assisted development (Claude)" in the stack line. In 2026 that reads as current, not as cheating — and it sets up an interview conversation you'll be prepared for (section 10).

09A GitHub profile that holds up to a click

Assume every interviewer clicks your GitHub for 90 seconds. Three things decide what they conclude:

• Pinned repos. Pin your three projects. Nothing else matters on the profile page.
• READMEs. Each project's README answers four questions in order: what is this, what's it built with, where's it live (link!), what would you add next. Five sentences is enough. A screenshot doubles the effect.
• Commit history. Thirty days of steady commits tells a truer story than any bullet point. Commit every working session with messages that say what changed ("add login redirect", not "update"). This happens automatically if you follow the plan — just don't dump everything in one commit on day 29.

10Interview prep: presenting AI-built work

The question you must be ready for: "So did AI write this, or did you?" The honest answer is also the strong answer:

Then let them ask. This only works if it's true — which is why days 22–30 of the plan push you to debug and modify, not just generate. An interviewer doesn't care that AI typed the code. They care whether you understand the thing you shipped.

Six questions to prepare for, per project

1. What does it do, in one sentence?
2. Why did you build it? (Your roommate story beats "for my portfolio.")
3. What was the hardest bug, and how did you fix it?
4. Walk me through what happens when a user clicks X.
5. What would break if 1,000 people used it tomorrow?
6. What would you build next on it, and why?

Write your answers down — a paragraph each, per project. Eighteen short paragraphs is your entire technical interview prep, and it's stronger than a month of grinding puzzle questions, because it's about work you actually did.

The mock interview checklist

• Do two full mock rounds out loud with a friend (or an AI assistant playing interviewer — ask it to be skeptical and interrupt).
• Practice the 90-second walkthrough of each project with the live URL open — screen share, not slides.
• Have one "it broke and here's how I found the problem" story ready. Debugging stories are the most credible thing a fresher can tell.
• Prepare one good question to ask them about their codebase. It signals you think like someone who ships.

11When you're stuck

• The 20-minute rule. Try on your own for 20 minutes. Then paste the exact error message and the relevant code into your AI assistant and ask what's wrong. Not "it doesn't work" — the exact error, the code, and what you expected to happen. Specific questions get working answers.
• Change one thing at a time. When debugging, make one change, test, repeat. Five changes at once means you no longer know which one mattered.
• Ship around it. Stuck two days on a feature? Cut the feature, ship the project, note it in the README under "what I'd add next." A live project with a known gap beats a dead perfect one.
• Don't switch projects. The urge to abandon and restart with a "better idea" hits everyone around day 15. The value is in finishing. Finish.

12If you want company doing this

Everything above works solo, and it's yours free — that was the deal. The honest difference a program makes is speed and accountability: someone who's shipped before reviewing your work daily, unsticking you in minutes instead of days, and making day 15 harder to quit.

That's what Build45 is: 45 days, live every evening, five people a batch, three deployed projects — the same philosophy as this playbook with mentors in the room. It includes interview preparation and placement support: portfolio review, mock interviews, resume help, and practice presenting AI-built work the way section 10 describes.

One thing we won't say: that we guarantee you a job or an income. Nobody honest can. What we can do is make sure that when you walk into an interview, you're holding real, live, working proof — and you know how to talk about it.

Questions about any step? Email contact@getmaxglobal.com — it reaches us directly.