I design, build and run self-hosted infrastructure, automation pipelines and AI tooling for teams who would rather not think about their servers. Docker-first, monitored, and still readable six months after launch.
Infrastructure and automation are the base. AI goes in when it removes real work, not because it looks good on a slide.
Self-hosted platforms built to survive updates, restarts and the person who inherits them.
Repetitive work turned into something that runs on its own and tells you when it doesn't.
Local models and retrieval pipelines wired into products, kept private and predictable.
A selection of systems I've built and operate. Automation-heavy, infrastructure-backed, and designed to grow without a rewrite.
A Pyrogram bot that watches authorised groups, calls custom API endpoints, fetches the content and caches the result — so the same file is never pulled twice. Permissions, rate limits and cleanup rules are built in so it behaves in busy groups.
Upload and delivery flow on S3-compatible storage with short-lived signed links. Repeat requests mint a fresh temporary URL against the stored object instead of downloading anything again.
Operational tooling for oversized assets: zip packaging, splitting when a file exceeds the limit, part tracking in a database, and cached re-delivery so a group never re-uploads the same archive.
An rclone-based backup routine tuned for large datasets, built around predictable runs, visible progress and as little daily friction as possible. Boring on purpose.
A self-hosted malware analysis platform that spins up throwaway VMs on demand, caps each session by time, and destroys everything afterwards. Currently in design and prototyping.
Delivery tracking with an admin panel, package management and GPS updates from drivers to estimate arrival windows. Web-first, with a backend structured for a real fleet rather than a demo.
Nothing here is exotic. Everything here is something I've operated long enough to know how it fails.
Five things I don't negotiate on.
Working across infrastructure and open data occasionally means running into things that shouldn't be reachable — an open storage bucket, a leaked credential set, an exposed admin panel. When that happens, I tell the owner. I don't do anything else.
That's the full extent of it. Whether or how they act on it is entirely theirs to decide — I don't monitor, chase, or follow up.
security.txt — and stays fully outside the
system the whole time.
Tell me what you're running and where it hurts. If it's something I can help with, you'll get a straight answer — and if it isn't, I'll tell you that too.