I build the whole system, not the part that demos well.
I take a product from an empty repository to something running on a server with its own domain, TLS, database, backups and a deploy that does not lose work. Backend, frontend, infrastructure, and the AI plumbing in between.
Everything below is live, and you can open it inside this page — those are the real sites, not screenshots. The chat is not an embed either: it runs on a model gateway I built, on a box I operate.
In production
Seven products, one operator
These share infrastructure I designed and run — one edge, one Postgres cluster, one cache, one object store, across three machines in two countries. Open any of them and it loads here, live.
Not an embed
Ask my model gateway
An OpenAI-compatible endpoint pooling several providers behind capability tiers. It balances on live load, benches a provider that fails twice, and falls forward instead of erroring. The same box is serving this page and that answer.
It draws with Stable Diffusion XL, reads its replies aloud, and keeps the conversation. Free to try, no account.
Open the chat →Capability