JD
EN/DE
Projects

Everything I've built,in one place.

The home page shows four case studies. This page has the rest: seven apps you can click through, smaller client projects, and a hands-on look at how I keep AI costs under control.

01Live on Vercel

Seven apps.One hardened stack.

My reference suite: seven end-to-end Next.js applications across hospitality, finance, retail, healthcare, trades and professional services. I designed, built and deployed each one, and every one is live. Click through and try them.

Shared stackNext.js 16React 19TypeScriptTailwind CSSPrismaBetter AuthPlaywright E2ECI gate
Screenshots from the live deployments.
02Client work

Smaller projects,real problems.

Short engagements from 2025, each solving one specific manual or brand problem for the client.

GRC client2025

Brand Template Studio

Editable HTML templates with in-browser PNG export, so a non-technical marketing team can produce on-brand assets without a designer.

  • Next.js 16
  • React 19
  • PNG export
0
designers needed per asset
FYOR · Fintech2025

Fintech Marketing Site

A static Astro + Tailwind build with GSAP and Lenis motion that ships no client-side framework JavaScript. Delivered with a written positioning audit.

  • Astro
  • Tailwind
  • GSAP
  • Lenis
0 KB
framework JS shipped
Healthcare & FMCG client2025

Product-Data Pipeline

Fragmented product catalogues consolidated into a validated marketplace upload workbook in Python, replacing manual data entry one SKU at a time.

  • Python
  • Validation
  • Marketplace upload
Batch
upload instead of per-SKU entry
03Interactive demo · simulation

AI that respectsa budget.

A scripted replay of the routing layer in my AI Content Studio. No model is called here. Each pipeline step goes to the cheapest model that can do the job well, and every token counts against a hard cap. Run it with smart routing, then switch to Opus-for-everything and watch the cap stop the job.

Illustrative rates and token counts. The routing and cap logic matches what I run in production.

content-studio · job_7f3aSSE
Spent$0.000 of $1.00
  1. Crawl & clean site
    Firecrawl · deterministic
    no LLM
  2. Classify pages
    short, structured
    haiku
  3. Extract brand voice
    nuanced reading
    sonnet
  4. Draft 12 social posts
    long-form generation
    sonnet
  5. Creative polish: hero post
    highest-stakes output
    opus
  6. Validate against schema
    Zod + cheap check
    haiku
$ await router.run(job, { cap: $1.00 })▍

Want something like this for your team?

Let's talk