I Built My Professional Landing Page Entirely with AI — Here's What Actually Worked
Most AI-generated content looks AI-generated. This one doesn't. Here's the process that made the difference.
I needed a professional web presence. I'm an engineer with twenty-five years of experience, not a designer, and I had no interest in spending weeks tweaking CSS or agonizing over copy. So I did what everyone does now — I asked AI to help.
The problem is that most AI-generated personal content sounds like AI-generated personal content. "Passionate technology leader with a proven track record of delivering scalable solutions." You know the voice. It's nobody's voice.
What I ended up with is lebedev.ai — an interactive, scrollable landing page that people genuinely cannot tell was AI-assisted. The feedback I've received consistently says the same thing: it sounds like a real person, not a template. This post is about how that happened.
The resume as starting material
It started with a resume. Not the landing page — a plain resume, the kind you send to recruiters. I went through seven versions before arriving at something that felt right.
The key insight came early: I asked for a version that reads like a non-native English speaker. Not with grammatical errors — with different word choices, more direct phrasing, slightly different idioms. I grew up in Russia, studied mathematics at Moscow State University, and have lived in the US for over twenty-five years. My English is fluent, but it is not native English. And that is part of who I am.
The difference was immediate. Compare the same section across versions:
The before version is competent. It says the right things. But it sounds like every other senior engineering resume. The after version has a point of view — "I did this on purpose, because working through the real problems myself was the fastest way to identify the true patterns." That is a specific person explaining a specific decision. It is not a template anyone could have written.
The feedback loop — and knowing when to push back
Every version went through external review. I collected feedback, evaluated it, and decided what to act on. This is the part most people skip, and it is the part that matters most.
Not all feedback is good feedback. One reviewer suggested making the headline "more declarative" without offering a concrete alternative. That is the kind of advice that leads to tinkering without improvement. I ignored it.
Another reviewer suggested adding a short skills and domains line after the summary. Two lines, no fluff, just enough for ATS parsing and recruiter scanning. That was specific, actionable, and low-cost. I added it.
The most useful feedback is specific enough to act on and honest enough to push back against. "Make it better" is not feedback. "This paragraph is redundant because the Experience section already covers it" — that is feedback.
Over seven versions, the changes were incremental but they accumulated. The summary went from four paragraphs to two. A "worth noting" hedge in front of the engagement survey data got dropped — the data was strong enough to stand on its own. The GenAI section gained concrete metrics (1.25M phone leads, 225K SMS leads) that transformed it from "we built production systems" to "we built production systems at this scale."
From resume to interactive primer
Once the resume felt right, the question became: how do you present this for the web? A PDF download is what everyone else does. I wanted something that demonstrated the kind of thinking the resume describes.
The concept I landed on is a "primer" — somewhere between a slide deck, a magazine article, and a small web app. It scrolls like a book, but it includes interactive elements: an animated career timeline, metric counters that tick up as you scroll, an architecture diagram, clickable panels that expand to reveal detail.
The aesthetic was intentional: warm paper tones, editorial typography (Libre Baskerville for headings, DM Sans for body), and a steel blue accent. The goal was something that feels designed, not generated. No purple gradients, no Inter font, no cookie-cutter layout.
The interactive elements are simple but effective. Scroll-triggered fade-in animations, hover effects on timeline items, counters that animate when they enter the viewport. All vanilla JavaScript — no framework, no build step for the static page.
// The core pattern: IntersectionObserver for scroll animations
const observer = new IntersectionObserver((entries) => {
entries.forEach(entry => {
if (entry.isIntersecting) entry.target.classList.add('visible');
});
}, { threshold: 0.15, rootMargin: '0px 0px -40px 0px' });
document.querySelectorAll('.fade-in').forEach(el => observer.observe(el));Simple, but it makes the page feel alive without being distracting. Each section appears as you scroll to it, which creates a natural reading rhythm.
The iteration that made it a real landing page
The first version of the primer opened with "The longer story" and described itself as "a companion to my resume." It was good content in the wrong frame. A reviewer pointed out that if this lives at its own domain, it should feel like a homepage, not an auxiliary page.
That one piece of feedback changed three things. The "longer story" label became a proper sticky navigation bar with anchor links. The "companion to my resume" intro was replaced with a standalone statement. And a prominent contact section was added — not buried at the footer, but as a styled call-to-action card you cannot miss.
Another reviewer flagged that the hero section took up too much vertical space — you had to scroll before seeing any hint that there was more content below. A simple CSS change (min-height: 80vh to min-height: auto) let the Philosophy section peek above the fold, solving the scroll cue problem.
These are small changes. But they are the difference between a document and a product.
What AI cannot do for you
AI wrote most of the code. AI generated the first drafts of every text section. AI helped iterate on voice, structure, and layout. But the decisions that made the final result good were all human.
AI cannot tell you what your career narrative is. It can help you articulate it once you know. The "IC to director and back again" framing — that was my story, my choice, my positioning. AI helped me express it clearly.
AI cannot decide what to cut. It over-includes by default. The third summary paragraph ("I have built search infrastructure, booking platforms...") was removed because the Experience section already covers it. AI generated it; a human reviewer flagged it as redundant; I agreed and cut it.
AI cannot judge whether something sounds like you. Only you know that. When a reviewer suggested softening the "managing around it" language because it might seem "anti-management," I kept it. That directness is part of my voice. A more diplomatic version would be more palatable and less honest.
Use AI as a skilled ghostwriter, not as an author. The ghostwriter produces drafts. The author decides what stays, what goes, and what the whole thing means.
The result
The site is live at lebedev.ai. It gets consistent feedback that it feels authentic, senior, and distinctly not generic. People browse it and do not think "AI made this." That, more than any technical achievement, is the test that matters.
The total process — resume through seven versions, primer, feedback cycles, deployment — took a few focused sessions. Not weeks. The AI acceleration was real. But the quality came from the iteration, the feedback, and the willingness to make editorial decisions that AI could not make on its own.
In the next post, I will write about adding an AI chat assistant to the site — a different kind of challenge, where the architecture decisions matter more than the prose.
