Realizing my old portfolio was stuck in time
A few days ago, I opened my old portfolio and felt that familiar mix of boredom and guilt. The structure still looked fine, but almost everything inside it was out of date. Roles had changed, projects had evolved, and the work I was proud of today lived in completely different places.
I could have sat down and edited every section by hand. But I knew how that story ends. You fix it once, it looks fresh for a few weeks, and then life moves on. Six months later, the page is stale again. It was not just out of date. It was built in a way that almost guaranteed it would keep falling behind.
At the same time, my day job as an AI engineer had been teaching me something very clearly: most clients do not just want clever models, they want automation. They want repetitive, fragile steps taken off their plate. That idea would not leave my head while I stared at my frozen portfolio.
So I asked a simple question: if I am building automation for everybody else, why is my own portfolio still manual?
Thinking like an automation engineer, not just a web designer
Once I put my “AI engineer” hat on, the problem looked different. The question was no longer “How do I rewrite this page” but “What are the sources of truth for the information here, and how can I connect them directly.”
I walked through each section:
Experience
I already keep my resume up to date. That is the one document I always touch when something important changes. Why should I also hand edit experience on the websiteProjects
My real project history lives on GitHub. Commits, repos, pinned projects, and descriptions already tell the story of what I have been building.Research & publications
Google Scholar knows my academic and research side. It already tracks citations, papers, and co‑authors.Content and hobbies
My newsletter holds my thoughts in writing. My YouTube channel holds my vlogs and experiments on video.
In other words, my portfolio did not need me to type things twice. It needed to listen to the tools I was already using every week.
That is when the design goal shifted from “make a new portfolio” to “build a portfolio that updates itself.”
Automating the core sections
With that mindset, I broke the site into parts and tied each part to a data source.
1. Experience from my resume
Keeping experience consistent across resume, LinkedIn, and portfolio is painful if you do it by hand. The reality is that the resume is usually the most accurate, because it is the one you tweak before interviews or applications.
So I decided:
The resume is the source of truth.
The portfolio should watch for a new resume and update itself.
The flow I designed in my head was simple:
I upload a fresh PDF resume to a specific folder in my cloud drive.
A small service picks it up, parses the experience section, and extracts roles, companies, dates, and descriptions.
The portfolio reads that structure and renders it into the experience section.
From my point of view as a human, “updating experience” becomes “drop new resume in the folder.” The website does the boring part.
2. Projects from GitHub
When you work in tech, your GitHub often tells your story better than any static paragraph. It shows what you are actually building and maintaining.
So instead of manually listing projects with screenshots that would age out, I wanted:
A Featured projects section, where I can pin or curate a handful of things I am especially proud of.
A Live feed of repos and contributions pulled from GitHub.
The idea is:
I mark certain repos as featured by following a naming rule, topic tag, or a small config file.
The portfolio fetches those and shows them with descriptions and tech stacks.
Everything else still appears in a broader projects area, so people can browse beyond the highlight reel.
Now, when I push a new project or update a readme, the portfolio quietly reflects that change. No more copy pasting descriptions into a CMS.
3. Research and publications from Google Scholar
The research side of my life evolves through papers, preprints, and citations. Google Scholar already tracks that far better than a manual list.
So this section became:
A link between my portfolio and my Scholar profile.
Automated listing of my latest or most cited work.
A simple way to surface research interests without rewriting them every time a new paper appears.
Again, the pattern repeats. I update the academic world in the place it naturally lives, and my portfolio listens.
Side projects and hobbies say a lot about a person, especially in AI and engineering. For me, that means:
Writing a newsletter about money, AI, and real life.
Creating vlogs and videos on YouTube.
These sections could easily become stale if I hard coded them. Instead, the site pulls:
The latest newsletter posts, so new readers always land on something fresh.
The latest YouTube videos, so visitors see what I have actually been talking about recently.
Every time I hit “publish” on a newsletter or upload a video, the portfolio has new material without me touching a single line of HTML.
At this point, the site starts to feel less like a static business card and more like a window into everything I am already doing elsewhere.
Adding an AI assistant on top of the data
Once the basics were automated, there was one more thing I really wanted to try. If my portfolio is now plugged into all these different sources, why should visitors be limited to clicking around section by section
That is where the idea of an AI assistant came in.
What I wanted it to do
I wanted the assistant to:
Help people navigate
If someone types “Show me your projects about agents” or “Where can I read your investing articles,” it should guide them to the right section or link.Understand my content, not only my layout
A lot of my thinking lives in my newsletter articles. If a visitor asks “How do you think about sleep at night investing” or “What is your view on F‑1 students investing in the US,” the assistant should answer based on what I have already written, not invent something completely different.Provide links back
Whenever it answers a question that comes from a specific article or project page, it should point the user back to the source so they can dig deeper.
How it changes the experience
The result feels very different from a static portfolio:
Someone who is in a hurry can ask one or two questions and jump straight to what they care about.
Someone who wants to understand how I think can talk with the assistant and get answers grounded in my own writing and work.
The site becomes not just a list of facts about me, but a searchable interface to my projects, research, and ideas.
For me, it is also a nice proof of concept. It shows the same philosophy I use with clients: connect real data, automate the tedious glue, and put an intelligent layer on top that helps humans find what they need faster.
Why this approach matters to me
Rebuilding my portfolio this way scratched several itches at once:
It matches what I do for a living
As an AI engineer, I spend a lot of time turning messy processes into smoother, automated flows. Now my portfolio behaves like that too.It respects my limited attention
I do not want to babysit a website. I want to update my resume, push to GitHub, publish a post, or upload a video, and trust that the portfolio will keep up.It tells a more honest story
Instead of a frozen snapshot from one good week, the site reflects my work as it is today. If I slow down, the site slows down. If I am busy building and writing, the site looks alive.
Most of all, it was a reminder that “portfolio” does not have to mean “static landing page.” It can be a living system that listens to the tools you already use, and an intelligent assistant that helps visitors explore what you are building and thinking about in real time.
If you are also in AI, engineering, or creative work, that is the core idea worth stealing. Before you redesign the front end of your personal site again, ask yourself: “What can I wire up so this thing mostly runs itself”