Build a portfolio that lands jobs

Recruiters skim. Hiring managers verify.

The single best move for an AI/ML resume is three honest, polished portfolio projects:

1. A classical ML project — train, evaluate, deploy. Show you know the loop.
2. A deep learning project — train a non-trivial model on a real dataset.
3. An LLM-powered app — RAG / agent / fine-tune. Show you can ship.

Each with a public GitHub repo, a short blog post, and a 90-second demo video. That's it. More projects don't beat three great ones.

Most ML repos have a README that says 'install requirements, run main.py'. Useless.

What I look for hiring:
- Problem statement (1 paragraph — what does this solve?)
- Demo gif / screenshot — show, don't tell
- Architecture diagram — your design, simply drawn
- How to run — copy-paste commands
- Results — numbers vs baseline, plots
- Trade-offs you considered + chose

10 minutes of README polish lifts a project from 'student' to 'engineer'.

Start every repo with this skeleton. Replace placeholders. Done in 30 minutes. Recruiters get the gist in 30 seconds.

Your GitHub profile is the first link recruiters click. They see your pinned repos.

By default, GitHub pins recently-updated. That includes hello-world forks. Bad signal.

Fix:
1. Profile → Customize your pins
2. Pick your three best (the portfolio above)
3. Make their READMEs polished

Add a profile README too — one paragraph + links. Spend 30 minutes. The signal upgrade is enormous.

Day 85 done.

- Three real projects > many tutorials
- README has 6 ingredients
- Reusable template
- Pin your best on GitHub

Tomorrow (Day 86): open source. The cheapest way to learn AT scale and get noticed.