Portfolio · Case studies · Backend & Infrastructure

Artem Mykhailichenko

Backend / Infrastructure Engineer

I build backend, infrastructure, CI/CD and production operations for systems where shipping is not enough: runtime evidence must prove the change works.

BackendInfrastructureCI/CDProduction operationsTelegram automationMedia pipelineLLM integrationDocker

Cases

Open a case to see the project link, backend ownership, service connections and delivery pipeline.

Project

Universe

Production Telegram Mini App and game platform

Project overview

Backend / Infrastructure Engineer

Universe is a production Telegram Mini App and game platform with economy, tournaments, raids, payouts, admin tooling, support tooling and live-ops processes. My focus was backend, infrastructure, CI/CD and production operations.

Backend · Infrastructure · CI/CD · Production operations

Product
Telegram Mini App

game platform with live-ops

Runtime
Kubernetes + Helm

staging / production rollout flow

Jobs
Celery + Redis

async game, payout and ops workloads

Ops
Evidence first

logs, DB state, SHA and health checks

System architecture

Services, state and integrations

01

Entrypoints and product surfaces

Telegram Mini App, bot, admin panel and support tooling converge into a FastAPI backend with shared domain rules and audit-visible mutations.

02

State and async workloads

PostgreSQL owns transactional state, Redis coordinates queues and Celery workers process game, payout, live-ops and operational jobs outside the request path.

03

External integrations

TON, TonConnect, Telegram, S3-like storage, Sentry and AI providers are isolated behind service boundaries with retries, logging and runtime verification.

Delivery & operations

How changes reached production safely

01

CI/CD ownership

Maintained GitHub Actions pipelines for backend, frontend, admin and support bot, including checks, image builds, registry pushes and rollout steps.

02

Runtime evidence loop

Verified changes through deployed SHA, health endpoints, logs, database state, Kubernetes rollout status and post-deploy behavior instead of assuming CI was enough.

03

Safe production changes

Used audit trail, idempotency markers, reread-after-write checks and delivery verification for production mutations and operational fixes.

Role

What I built and operated

Built backend features and service-level business logic.
Investigated live and staging incidents using a runtime evidence first approach.
Diagnosed issues through database state, logs, Kubernetes, GitHub Actions, health endpoints and deployed SHA.
Maintained CI/CD pipelines for backend, frontend, admin, support bot and deployment flows.
Worked with Alembic migrations, Postgres, Redis, Celery, Telegram Bot API, TonConnect / TON and S3-like storage.
Performed safe production mutations with audit trail, idempotency markers, reread-after-write and delivery verification.
Carried changes through PR, CI, staging / production rollout and runtime verification.

Engineering decisions

Trade-offs behind the backend design

01

PostgreSQL as source of truth

Transactional game, user, economy and payout state stays in PostgreSQL, while Redis remains coordination infrastructure, not permanent truth.

02

Async work outside requests

Celery separates slow or failure-prone workloads from user-facing API paths, making retries and operational visibility explicit.

03

Deployability over local success

Changes were carried through CI, image build, staged rollout and runtime checks so the actual deployed system, not only local code, was validated.

Stack

Backend, infrastructure and delivery stack

Backend

Python 3.12+ · FastAPI · SQLAlchemy · Alembic · Pydantic · Celery · Redis · PostgreSQL · asyncpg · aiogram · httpx · Sentry · Prometheus

Infrastructure

Docker · Docker Compose · Kubernetes · Helm · CloudNativePG · KEDA / ScaledObject · Sealed Secrets · cert-manager · Traefik / Ingress · GHCR · GitHub Actions · self-hosted runners / ARC

Quality & Security

uv · ruff · mypy · pytest · pytest-xdist · bandit · pip-audit · migration checks · CI artifacts · /livez /readyz · audit logging · idempotent operations

Architecture maps

Runtime architecture

System map

Runtime architecture

Runtime flow

01

Entry points

Telegram User
player entry
Mini App
React / Vite
Bot + Support
Telegram API
Admin Panel
operations
02

Application core

FastAPI Backend
business logic
Celery Workers
async workloads
Celery Beat
scheduled jobs
03

State + integrations

PostgreSQL
CloudNativePG
Redis
queues / cache
S3 Storage
objects
TON / AI / Sentry
external services
04

Delivery layer

GitHub Actions
CI/CD
GHCR
images
Helm
release
Kubernetes
runtime