AI engineer portfolio: what hiring managers actually look at
A strong AI engineer portfolio has three to five deployed projects that solve real problems, each with a README that covers the problem, the architecture, how you evaluated it, what broke and how you fixed it, and a live link. Add a GitHub profile README and a simple portfolio site that link everything together, plus write-ups or short videos about your builds. One project with depth in a domain is worth more than twenty tutorial clones, and if showing a project would make you cringe, it isn't ready.
Why most portfolios get skipped
Recruiters and clients see the same chatbot-over-a-PDF project again and again, often running only on the candidate's laptop. A project on your laptop needs you in the room to show it. A deployed project with a URL or an API endpoint can be shared with a recruiter, a client or anyone, and it also shows you understand deployment, which every company needs eventually.
What each project should show
- A real problem, ideally in a domain you know or a domain the company works in.
- Your own data, messy enough that you had to make real parsing and retrieval decisions.
- An architecture diagram.
- How you measured quality, and the numbers.
- What broke, what you changed, and how you checked the change worked.
- Deployment: a link, a container, or a package someone can install.
- Limitations you know about.
README order
Most people read the first screen and leave, so put what proves the most first.
1 - Problem, in one or two sentences, and who has it. 2 - Demo link or a short video. 3 - Architecture diagram and the main decisions. 4 - Evals: what you measured and the results. 5 - What broke and how you fixed it. 6 - Limitations. 7 - How to run it.
Signature project
Pick one project to be your signature project and go deep: a video walkthrough, a write-up of the problems you hit, and the story of how the design changed. People respond to stories, so explaining how you reached the solution matters more than listing the tools you used.
GitHub profile README
Create a repository with the same name as your GitHub username and add a README.md, and GitHub shows it at the top of your profile. Use it as a landing page: what problem you're working on, what you're building now, writing you've published, tools you've used, and a short background.
Portfolio site
A domain is cheap and hosting a static site is free on platforms like Vercel, so a simple site is worth it, especially if you're moving toward leadership. You don't need to design it from scratch: start from a portfolio template (Vercel and v0 have many), push it to your own repository, and edit it with a coding agent like Claude Code.
| Section | What goes in it |
|---|---|
| Home | One line on the problems you work on |
| Projects | Your signature project first, with links |
| Blog | What you think: explanations, opinions, lessons |
| Build logs | What you're working on now, in short updates |
| About | Background and how to contact you |
Engineer and leader portfolios differ
| Engineer | Leader |
|---|---|
| Deployed projects and code | Case studies: situation, decision, result |
| Evals and what broke | Evaluation frameworks and how adoption was measured |
| Open source contributions | Talks, workshops and training you've run |
| Build logs | A vision statement and adoption playbooks |
| Technical write-ups | Writing on how you think about AI decisions |
Leaders usually can't show code from their company. What they can show is a case study: what the situation was, what you tried, what didn't work, what you changed, and what happened.
Turn what you already know into content
You don't need to invent topics to write about. Record a 10 or 20 minute talk on something you're working on, transcribe it, and turn the transcript into a blog post, because the knowledge you and your organisation have is something nobody else has.
Certifications
Most certifications add little to an AI engineering portfolio. Cloud certifications are the exception for engineering roles, because they show you can work with the platforms companies deploy on. Before you spend time and money on one, check whether companies hiring for your role value it, and what you'll learn from it.
Target the company
Instead of sending the same resume to a thousand companies, look at what problems a company you want to join is solving, build a project around that problem, and show it to them.
From the floor
In the portfolio session for AI Engineer HQ, I showed builders how to start from a v0 template, publish it to their own GitHub repository, clone it and keep editing with Claude Code, and I separated what engineers should show (what they built) from what leaders should add (case studies, evaluation frameworks, talks, adoption playbooks). I also said that certifications, for the most part, nobody cares about, and that the unique thing about you over the next few years is your own knowledge, so document what you work on, because you can turn transcripts of your sessions into blogs and other content.
In a cohort build session, I described the cringe test we use for every artifact: if someone asked what's in your portfolio and you showed this project, would you be proud, or would you cringe because it doesn't have enough depth? I said one of our own artifacts wasn't at that level yet given my years of experience, so we'd keep improving it until it passes.
Related
- Roadmap
- Git and GitHub
- AI engineer resume
- Learn by building