A portfolio built like a product

Hiring managers do not hire screenshots. They hire people who can frame problems, ship with quality bars, and operate responsibly when tools and models move fast. A static portfolio optimises for the first impression; it rarely survives the second question: *How does this person work with engineers? How do they scale content? How do they prevent AI from becoming an expensive liability?*

I needed a site that could stand up to that scrutiny: multi-locale reach without fragmenting the experience, a coherent design system that stays on-brand, and an AI feature that is grounded—not generic chat wrapped in marketing.

The constraint was intentional: no CMS team, no endless bespoke pages—just disciplined content models, automation where it earns its keep, and review where judgment matters.

Design craft. The interface is built on shadcn/ui (Maia) so components stay consistent, accessible, and theme-native—no one-off "fix it in CSS" drift. Layout and hierarchy follow a tight spacing scale and Apple-style interaction principles baked into how the work is executed, not just how it is described. The result is a portfolio that feels intentionally designed, not assembled.

Product architecture. Next.js 16 and the App Router give a clear split between what can be static, what must be dynamic, and what should stay server-side. next-intl carries the same information architecture across English, Hindi, Arabic, and Spanish—a deliberate statement about inclusive reach and execution depth, not a plugin checkbox.

Tooling as operating model. GitHub is where quality becomes habit: pull requests, automated i18n linting, and workflows that keep product documentation and issues aligned with what actually ships. Vercel is the deployment layer—fast previews, production promotion, analytics and performance signals—so iteration stays tight and regressions surface early.

AI, treated like a product feature. The Ask experience is backed by retrieval: content is chunked, embedded, and seeded into Neon Postgres with pgvector so answers trace back to real sources. The build pipeline is careful about cost and safety—for example, machine translation and other API-heavy steps are not wired into default production builds; generated translations live in the repo and are reviewed. That is the same instinct you want in a production org: automation with brakes.

What hiring managers get is not "I used AI." It is evidence of end-to-end ownership: from interaction and visual design through internationalisation, content operations, search and structured data, and safe AI delivery. The site is faster to evolve because the system is modular; it is safer to operate because guardrails are explicit; it is more useful to visitors because Ask turns a passive archive into an interactive brief on how I work.

If you are evaluating a product designer who partners like a PM and ships like someone who respects engineering reality—this project is the working proof.