A working AI portfolio
Ten AI products. Live and running.
Each one is a real, deployed app. I built them end to end, from the model and the cloud infrastructure to the interface. Open any demo and use it yourself.
- 10 live products
- Built end to end, solo
- Running on production GCP
How I build
How each product gets built.
The same path for all ten: pick a real workflow, design it, build a live demo, then ship it to production.
Pick the workflow
Each product starts from a real small-business pain: booking, lapsed customers, daily care reports. I pick the one workflow worth automating and scope it tightly.
Design it
A short design doc first: what the AI does, what it sounds like, and what data it touches. The build stays sharp because the thinking happens before the code.
Build the demo
A live, working version on realistic data and a real brand voice, with AI output that streams in. Something you can click through, not a mockup.
Ship it
Polished and deployed to its own subdomain and Cloud Run service on production GCP. All ten are live right now, each one open to try.
Under the hood
Real AI, real infrastructure.
Real AI, not wrappers.
Multi-turn agents, voice to structured data, RAG over private docs, and multimodal report generation. Streaming live, powered by Gemini 2.5 and Vertex AI.
Production infrastructure.
Every app runs on Google Cloud Run with Firestore, native vector search, and path-filtered CI/CD. Real deployments on real infrastructure, not localhost screenshots.
Product and design, end to end.
A shared design system, accessible and mobile-perfect at 375px, with skeleton loaders and subtle motion. Built solo, from the model to the pixel.
Good to know
Questions you might have.
Get in touch
Take a look around.
Then let's talk.
Start with the demos. Every one is live and ready to try. If you would like to talk through the work, the architecture, or the code, I am one email away.