Ryan Ilano, AI-Native Design Engineer, New York.
I design and build. No handoff, no translation layer: I take work from Figma to HTML/CSS to production. Today that means AI products: coding agents on real work, routed across local models, DeepSeek, and frontier APIs by what each task actually needs. I call the split Gate and Wave. At Hearst the same practice ran at scale — one design system across 85 magazine sites and 60+ brands, on a network Hearst credits with 550M website visitors; an automated Apple News pipeline behind an estimated $10–20M/yr revenue line; and design lead on the accessibility program that reached 95%+ WCAG 2.1 AA (Deque Axe), with 120+ properties audited program-wide. Meticulous about style. Comfortable in a diff.
Projects
Selected Works-
Subfolio: AI Porting AREA 17's Filesystem CMS Design Engineer · Solo port -
Apple News+ & Hearst Magazines Design Lead -
mediaOS: A Unified Design System Design Systems & Platform Lead · Hearst Magazines -
Scaling Accessible Design Systems Accessibility Design Lead · Hearst Magazines -
Hearst Magazines: Custom Feeds Redesign Design Lead · Hearst Magazines -
Output: Sundays on The Roof Art Direction & Design, Branding. Print Digital and Hardgoods -
K2 Snowboarding Boots Lead Designer · Noct Bureau -
The Corner: NYC Techno Record Label Design & Art Direction
Writing
Deep dives for the technical reader-
The Bottle on the Counter A prescription label went up in a photo I uploaded for an unrelated reason. Post-hoc filtering leaves you trusting a retention policy. Pre-upload stripping leaves nothing to trust, because nothing moved. So I built the tool that way. -
...Endtroducing A little background on me and my work, and a long-form version of my about page. -
Resurrecting Subfolio: Bringing a Ten Year Old Tool Back with an LLM and Elbow Grease AREA 17's Subfolio CMS ran on PHP 5.6 EOL and was broken. Here's how I ported it to AstroJS and made it free to run on Cloudflare and GitHub Pages. -
Local Self-hosted AI is now Useful with 16GB of VRAM Open weights got good, which is the only reason any of this plumbing is worth doing. Meticulous notes on setting up a Proxmox LXC with NVIDIA GPU passthrough, CUDA, and llama.cpp. -
How I Run a Claude and DeepSeek Multi-Agent Dev Pipeline I learned agent orchestration by doing it by hand: Opus as architect and verifier, DeepSeek on the volume work, and a $20 Claude plan that goes a lot further than it should.
I'm not against vibe code. I'm against vibe code in prod.
If that distinction matters to you, we should talk.