Not a portfolio wall. Three real systems, each scoped, built, deployed, and being used by real teams today. Here's the honest version of each: the problem, what we built, and what changed.
A capable engineering team was losing hours to internal overhead, standups, timesheet review, sifting through logs, tracking who was available. Useful work, but the kind that quietly eats a team's week.
A self-hosted platform of five AI agents, each handling one of those chores. Single sign-on, answers grounded in the company's own internal knowledge base, deployed as three containers on infrastructure they already owned.
The agents run in production today. The team gets its standup summaries, timesheet checks, and log triage without a person doing them by hand, and every action the AI takes is recorded and reviewable.
Small businesses run on WhatsApp, but the work behind it is all manual: customers asking the same questions, orders in a notebook, payments confirmed by squinting at screenshots. It works until it doesn't scale.
A multi-tenant platform that handles customer conversations, confirms payments, tracks orders, and hands over to a real person the moment the AI shouldn't be the one deciding. Built on a self-hosted automation and chat stack.
Live with pilot businesses now, onboarding more. Owners get their evenings back while the platform handles the routine, and the handover-to-human step keeps customers from ever feeling stuck with a bot.
Which client is signed up for which service, and how each one is set up, lived in scattered spreadsheets. Different people kept different copies, so no one could be sure what a given client actually had, and a large service migration that depended on that record was fragile and easy to get out of sync.
A self-hosted registry where the product catalogue is stored as data, not code, so non-technical staff can add new services and configuration themselves without a developer. A simple grid shows who has what; a flexible detail layer captures the specifics (versions, sub-types, and the hundreds of fine-grained items a single service can carry).
Becoming the single source of truth for client subscriptions, and the place a major migration is planned and tracked from. Clean dropdowns and tags instead of free-text keep the data trustworthy, with an audit trail and export built in.
If something in your operation is quietly broken, or quietly eating your team's week, tell us about it. The first conversation is free and honest.