The studio site: two languages and eight WebGL scenes held to one performance budget.

Role
Founder and CEO. Design and engineering.
Studio
Kavaliero
Year
2026
Built with
Next.js 16, React 19, TypeScript, Tailwind CSS 4, GSAP, Lenis, three.js, next-intl
Kavaliero home page. The headline reads: Design that moves your company forward.

Kavaliero is the studio I founded in 2026. It sells three things in a fixed order: websites, SaaS platforms and AI automation, first to companies in Venezuela, then Latin America, then everywhere else.

I designed and built the site in Spanish and English. It is the studio's first case study, so every effect on it is one we are prepared to build for a client.

  • 32localized URLs
  • 8WebGL scenes, all code-split
  • 2languages, each written natively
  • 160 msblocking on arrival, down from 1.7 s
Kavaliero launch film, Spanish cut, 26 seconds.

Motion

  • One wheel, four screens

    The hero and the three service screens sit on an edge-on wheel driven by a single position function, so the opening scrubs as one object instead of four sections.

  • A pixel curtain between pages

    Internal links are intercepted. The page covers in 0.42 s, reveals in 0.54 s, and arrival animations wait until the curtain has cleared.

  • A preloader that preloads

    The logo is sampled into cells on a canvas while the 3D models prefetch behind it, with a CSS-only failsafe if scripting stalls.

The home page, scrolled. 12 seconds.

WebGL

  • Procedural ice

    The crystals on the Work page have no model files. Geometry, textures and a transmission material are generated in code and built across several frames.

  • One budget for eight scenes

    Every scene is code-split, paused off screen and in hidden tabs, capped on pixel ratio and given a reduced-motion fallback.

  • A measured freeze fix

    Compiling shaders in parallel cut main-thread blocking on arrival from 1.7 s to about 160 ms on the reference machine.

Kavaliero work page with projects sealed in blocks of procedural ice.
Automation page hero: Automating is the road to success, over streaking road lights.

Under the surface

  • Routing that reads the reader

    Spanish is unprefixed and English lives at /en. Slugs are written natively in each language and paired for hreflang.

  • Content as typed data

    Articles are TypeScript modules. An answer block feeds FAQ structured data, and a stat block is not valid without a source link.

  • Closed by default

    Strict content security policy, HSTS preload, and a contact form that validates on the server, traps bots with a honeypot and fails closed without a mail key.

About page: Born in Venezuela. Forged far away from home.
A white robot leaning into frame in front of oversized type.

NextKeep scrolling

Lead search, qualification and outreach in one workspace, with a person approving every send.

Role
Product, interface system, front end and AI flows.
Studio
Kavaliero
Year
2026
Built with
React, TypeScript, Vite, Tailwind CSS, Node.js, Express, Prisma, PostgreSQL, OpenAI API, Stripe, Puppeteer
Live
Case study at kavaliero.com
Leadesse lead search: keywords, location, radius and AI strictness beside a map of results.