Computer Science student at Nazarbayev University. I build backends that are deployed and tested: payments with Stripe webhooks, real-time chat across server instances, background-job platforms and e-commerce APIs — Python/FastAPI and Node/TypeScript on PostgreSQL and Redis. I work with AI-assisted tooling every day.
Available 20–25 h/week now, full-time in summerLanguages Kazakh, Russian (native) · English (C1)Open to junior roles, internships, freelance
Every project below is live — the line under each name is a real endpoint you can open.
Paygate
Payments, credits and an AI assistant that fails gracefully
GET /health 200
Users buy credit packs through Stripe Checkout; a signature-verified webhook credits a ledger; credits are spent on an OpenAI-backed assistant. Idempotency keys, event-id dedupe and row locks mean a double-delivered webhook can never credit twice. Circuit breakers, retries with backoff and rate limits wrap every outbound call.
Rooms, presence, typing indicators, delivery and read receipts — working across several server instances behind nginx via a Redis Streams adapter, with connection-state recovery so a dropped socket loses nothing. A two-instance integration test runs in CI.
PDF reports, bulk email and image compression run as queued jobs that answer 202 with a job id. Exponential-backoff retries, a replayable dead-letter queue, soft and hard time limits, stuck-job recovery and tag-based cache invalidation.
Rotating refresh tokens with reuse detection, role-based access, checkout that deducts stock inside a row-locked transaction, Postgres full-text search and integration tests against a real database — deployed as one Vercel project with a serverless FastAPI function.
Guests open a link or a QR code, order and pick up without queuing. Staff get a password-protected admin with a stop-list; the owner gets a dashboard and a one-tap WhatsApp daily report. Built for and pitched to a two-location coffee chain in Astana.