I build systems that scale,
end to end.
Infrastructure, security, backend, frontend.
Built and run on infrastructure I operate myself.
Main projects
letterhop
Microsoft is retiring SMTP basic auth. The day it goes, every printer, scanner and ERP box that has been mailing for a decade falls silent. letterhop sits in between: a relay on the customer's network takes classic SMTP from the device and delivers through OAuth2 to Microsoft 365, or through an ordinary mail server. Built for managed service providers running many customer sites.
- Accepts mail on 587 STARTTLS, 465 TLS and, only when enabled, plain port 25 – whatever the old device still speaks
- Writes every message to disk before it answers 250 OK, then retries and dead-letters – a scan is never silently lost
- Switches to an upload session by itself when a scan is too large for a normal send
- Installs as a Windows service; the multi-tenant cloud manages the fleet, and mail never passes through it
- An enrolled relay always delivers: a lapsed licence locks its configuration, never the mail
- Signed licences, single-use enrolment tokens, TOTP and Cloudflare Turnstile on the platform
Background
I spent two years as an IT field technician and two more as an IT consultant at a managed service provider. I ran Linux servers, built homelabs, and learned what happens when infrastructure fails at 2am.
That background is why I design systems the way I do – I know what runs underneath them. For the past three years I've been building tools and analysing architectures; for the last year, large-scale systems like Dylaris.
How I work
I build everything with Claude Code. I don't write code by hand and I don't read it line by line. I'm the architect; Claude writes the code. I work from the explanation of the code, not from the code itself.
What that leaves to me is most of the engineering:
It filters in both directions. Teams that hire through live-coding whiteboards won't find what they test for here. Teams that build AI-first will find someone who already works that way.
Stack & capabilities
Go on the backend, TypeScript on the front. Everything below runs in production somewhere, not just in a sandbox.
Other projects
LabExtend
Swarm-native infrastructure monitoring as a single Go binary that runs as a global service. Source on github.com/Bartis-Dev/LabExtend.
- One node becomes leader and serves the UI, the rest run as agents over bidirectional gRPC
- Container logs stream live from a SQLite ring buffer, five thousand lines per container
- Alerts carry hysteresis, so a flapping container does not become a wall of notifications
- Restart-loop detection: three restarts inside sixty seconds
- Accounts on the large video and streaming platforms connect over OAuth2, each with its own token lifetime
- Background workers keep statistics fresh without the creator touching anything
- Agency-facing sharing with magic links and password-protected profiles
- Drag-and-drop kit builder on the front, rate limiting at the edge of the API
Creator analytics platform
Verified media kits for creators – the numbers on a profile were fetched by the platform, not pasted in as screenshots.
reclaim
A voxel survival game, and the one project here that is not a web service. The only thing I have built where the budget is frame time rather than requests.
- Procedural world generation: noise fields for forests and meadows, rule-grown trees, ore-bearing rock
- Voxel mining with the core written as a GDExtension in C++
- Multiplayer that works across two real processes, not two windows of one
- The overlay reports 1% lows rather than an average frame rate, because the average hides the stutter
Work together
Got a project that needs solid backend work – or a complete product, from the servers up?
Backend systems, distributed architecture, full product delivery – or a tool that just needs to actually perform. Open to projects and to a permanent role. Let's talk.