About my work.
I architect and deliver headless digital storefronts with Next.js, handling everything from UX design to cloud deployment.
MiniDist: Rebuilding a Pipeline That Was Silently Failing..
Dsp-pipeline started as a real problem at Alter K Global Music Services, where artists would upload their music and simply never hear back. No error. No notification. Just silence.
The original system worked when we were small. As volumes grew, large files timed out on upload, DSP API failures went undetected, and platform-specific logic was scattered across the codebase. I rebuilt it from scratch as three clean layers: S3 multipart upload to eliminate timeouts on files up to 200 MB, a BullMQ worker pipeline that handles encoding asynchronously without blocking the API, and a DSP adapter layer that isolates each platform — Spotify, Apple Music, Deezer — behind its own interface.
This project is a portable rebuild of that architecture, with a live demo that walks through the full pipeline in real time.
Live demo • GitHub
Shopify Flow:
Building the Headless Connection.
Atelier started as a custom e‑commerce platform and became a playground to integrate Shopify into a headless architecture. I worked with both the Storefront GraphQL API and the Admin API to experiment with importing and synchronizing product data into my own Node/Express backend.
The goal was not just to display data, but to connect a fully custom Next.js storefront—with advanced product filtering, URL‑synced state and SSR for SEO—to Shopify as a complementary commerce engine. On top of that, I integrated Stripe Checkout and built an internal admin area to manage products, stock and pricing on the custom side, while keeping the door open for deeper Shopify integration.
This project sits at the intersection of two worlds: the flexibility of a bespoke full‑stack setup and the reliability of a global commerce platform.
AI Image Playground: Combining Images with Multimodal AI.
AI Image Playground combines two images using Google Gemini 2.5 Flash Image with a textual prompt, built on Next.js 14, React 19, and TypeScript. The architecture leverages Vercel AI SDK for seamless integration, with AI Gateway providing monitoring and cost control. The application features drag-and-drop uploads, real-time previews, and optimized image processing that converts formats automatically. Designed for scalability, it uses serverless functions for horizontal scaling and maintains a stateless architecture for maximum flexibility. The system balances performance with observability, offering complete visibility into AI usage while ensuring efficient resource consumption.
The frontend handles image uploads and validation, while the backend processes images through format conversion before sending them to Gemini. The multimodal model processes both images and text in a single request, returning generated images as base64 data. This architecture represents an optimal balance between simplicity, performance, and scalability, ready for growth without major refactoring.
Let's build something better.
I strongly believe that designers and developers have a responsibility to make sure that what we are building does no harm and I try to be as ethical as I can in taking on projects.
If your organisation represents online gambling, payday loans, big tobacco, or mines and monetises personal data, then I am probably not the best fit for your project. Due to its extremely wasteful energy consumption, I don't take on projects using crypto technology. I'm not one to completely write off a technology but I can't conscionably work with it in its current form.
All that said, if you are looking for help building something that promotes sustainability, diversity, or generally aims to make a positive impact, then let's talk.