Backend engineer — Python · C#/.NET
Backends
that stay up
I design, ship and run production systems. Five of them are live right now, and I'm on call for every one.
Open to backend roles & contract work
Backends
Status ● nominal
Users 5,900
T+ 00:00:00
01 — Media fleet Live
Five bots.
One server.
Media bots on a Telegram-compatible messenger. Every file is cached by ID, and duplicate requests share one download.
5,900Users
27KFiles / week
0.48SP50 cache hit
Try to take it down.
A live model of how the fleet handles failure. Kill a worker, deploy mid-traffic, go viral. Break something.
0Delivered
0Requests lost
0Charged for failures
–SCache hit p50
–SPaid wait now
Requests are flowing. Green dots are paid users, white are free. Pick something to break.
A simulation of the production design: same queues, coalescing, janitor, watchdog, bounded drain and charge-on-delivery rules, with timings scaled to production medians. The counters above are from this simulation, not from production.
02 — Wallet · In production
Every balance equals its postings.
A double-entry ledger that reads bank notifications and credits a matching transfer with no human in the loop. Card numbers are kept only as HMACs, and every write is idempotent.
0Suspense
0.5SBank → credit
1Command to replay
03 — Fleet control plane · In production
One command.
Any server.
A primary holds all state and pushes config to stateless replicas: validate, swap, health-check, roll back automatically. Swapping a server never changes a user's link.
400Days of usage data
5MINAccounting interval
0Link changes on swap
04 — Also shipped
More launches
Message pipeline
One codebase, five messengers. Each platform is just an adapter.
Client · prodnobitex-peach
Exchange API client: REST, Ed25519 auth, async WebSocket order books.
PyPIproapi
Layered ASP.NET Core 10 API template, 27 tests out of the box.
C# · .NETbsky
Self-hosted crossposter from Bluesky to X and Instagram.
Open sourceMessage-Board
Real-time chat rooms. No sign-up, nothing to leak.
Open source