ACCESS GRANTED ›

// Subject Classification: Backend Systems Engineer

SAYOMIYORI

Python · Backend · Systems

$ stack FastAPI · Django · PostgreSQL · Redis · Docker
$ patterns Microservices · Event-driven · CI/CD · Observability
$ brokers RabbitMQ · Redis Pub/Sub · Celery
$ status OPEN TO OPPORTUNITIES ▋
Scroll
0 Repositories
0 Commercial Orders
0 Deployed APIs
0 Kwork Rating
Subject Dossier
Classification // Developer Profile ● Online
Sayomiyori
Backend Engineer
FunctionPython Backend Developer
LocationSimferopol / Remote
EducationCrimean Federal University
Projects7 repositories
StatusOpen to opportunities
// Primary Focus Areas
Microservices — Django + FastAPI + RabbitMQ
Event-driven architecture & message brokers
CI/CD pipelines, Docker, Observability
WebSockets & real-time systems
Clean Architecture & production-ready code
Tech Stack
// Backend
Python 3.12
FastAPI
Django / DRF
Celery
asyncio
// Data
PostgreSQL
Redis
RabbitMQ
SQLAlchemy 2
Alembic
// Infra
Docker
Nginx
GitHub Actions
Linux
Railway
// Quality
pytest
Prometheus
Grafana
Sentry
structlog
Project Files
File #001 // MicroserviceDEPLOYED
TaskFlow
Event-driven task platform — Django + FastAPI + RabbitMQ
DjangoDRFFastAPIRabbitMQChannelsNginx
▸2 microservices via RabbitMQ topic exchange
▸Real-time WebSocket (Django Channels + Redis)
▸JWT, role-based perms, Nginx reverse proxy
File #002 // SaaSDEPLOYED
WebHook Manager
Reliable webhook delivery with circuit breaker pattern
FastAPICeleryRedisPrometheusSentry
▸Exponential backoff (10s→1h), HMAC verification
▸Redis circuit breaker — zero event loss
▸Clean Arch, >80% coverage, auto-deploy
File #003 // Real-timeDEPLOYED
RealTimeChat
WebSocket chat — horizontal scaling via Redis Pub/Sub
FastAPIWebSocketRedis Pub/SubJWT
▸Fan-out across N replicas via Redis channels
▸Rooms, typing indicators, history on connect
▸Fully async (asyncpg + redis.asyncio)
File #004 // REST APIDEPLOYED
BookFinder API
REST API for book search with full monitoring stack
FastAPIPostgreSQLPrometheusGrafana
▸10 endpoints, JWT, Google Books integration
▸Prometheus + Grafana: error rate, p95, RPS
▸CI/CD: ruff → pytest 87% → Docker → deploy
File #005 // CommercialDELIVERED
ChillLibrary
Telegram bot — 3 AI-powered book search modes
aiogramGoogle APIsGemini AIVision OCR
▸Search by title, cover photo, or quote
▸Full cycle: spec → deploy → docs
▸Accepted without revisions
File #006 // CommercialDELIVERED
BodyTelling
Fitness bot — workout selection & retention mechanics
aiogramAPSchedulerpytest
▸Algorithm: 5-parameter workout selection
▸Streaks, achievements, TZ-aware reminders
▸Full docs, production handoff
File #007 // ETL PipelineDEPLOYED
EventPipe
Microservice ETL pipeline — event ingestion, processing & dual storage
FastAPIKafkagRPCMinIOKubernetesGrafana
▸REST + gRPC ingestion via Apache Kafka broker
▸Transform service: validation, enrichment, normalization
▸Dual storage: PostgreSQL + MinIO with pre-signed URLs
File #008 // AI PlatformDEPLOYED
AgentHub
AI platform — RAG pipeline + agentic tools + multi-provider LLM
FastAPIpgvectorGeminiAnthropicRedisMCP
▸RAG with semantic search, reranking & MCP server
▸JSON-based agent tool protocol: KB queries + web search
▸Semantic cache via Redis cosine similarity (≥0.95)
File #009 // Auth ServiceDEPLOYED
AuthFortress
Production-ready auth microservice — OAuth2, 2FA, RBAC
FastAPIPostgreSQLRedisOAuth2TOTPDocker
▸JWT rotation + Google / GitHub / Yandex OAuth2
▸TOTP 2FA, RBAC, audit logging, rate limiting
▸CI/CD via GitHub Actions, fully containerised
File #010 // ObservabilityDEPLOYED
PipeWatch
Real-time log aggregation — live tail, alerting & Rich CLI
FastAPIClickHouseRedis StreamsWebSocketPrometheus
▸Log ingestion + WebSocket live tail streaming
▸Alert rules with Telegram notifications
▸CLI client (Rich UI) + 30-day bloom filter indexes
File #011 // ML PlatformDEPLOYED
NeuroClassifier
ML inference platform — LoRA fine-tuning & ONNX Runtime serving
FastAPIPyTorchONNXHuggingFaceMinIOCelery
▸Async LoRA fine-tuning — trains ~1% of model params
▸3-5x faster inference via ONNX Runtime vs PyTorch
▸Model versioning & rollback via MinIO storage
Architecture Log
TaskFlow
Django for ORM-heavy domain logic + FastAPI for async notification consumer — each framework where it's strongest. RabbitMQ topic exchange decouples writes from delivery.
WebHook Manager
Celery + Redis broker for fire-and-forget delivery. Circuit breaker prevents hammering dead endpoints. HMAC + idempotency at infra layer, not business logic.
RealTimeChat
Redis Pub/Sub over in-memory — scaling to N replicas is transparent. WS auth via query param because browsers don't support custom headers on handshake.
BookFinder API
Async SQLAlchemy 2 + asyncpg — bottleneck is I/O. Prometheus from day one — errors visible in dashboard before the first bug report.
EventPipe
Kafka as the single source of truth for event ordering. gRPC for internal high-throughput ingestion, REST for external clients. MinIO stores raw events — PostgreSQL stores only enriched, queryable state.
AgentHub
pgvector over dedicated vector DBs — keeps the stack simple while supporting semantic search. Redis semantic cache at cosine ≥0.95 cuts LLM calls on repeated queries. MCP server exposes tools to any compatible agent.
AuthFortress
Token rotation on every refresh — stolen refresh tokens expire on next use. TOTP secret stored encrypted; audit log is append-only. Rate limiting at Redis layer, not application layer — survives restarts.
PipeWatch
ClickHouse for logs — columnar storage gives 10-100x faster aggregations vs PostgreSQL at scale. Redis Streams as ingestion buffer absorb bursts before ClickHouse write. WebSocket live tail is read-only — no polling.
NeuroClassifier
LoRA trains only 1% of parameters — fine-tuning on consumer hardware in minutes. ONNX export after training removes PyTorch runtime from serving path — 3-5x latency improvement. MinIO versioning enables instant model rollback.
Kwork Orders
★★★★★ 5.0 / 5.0
Отличная работа! Разработал Telegram-бота с нуля — чистый код, документация, деплой. Сдал раньше срока, на все вопросы отвечал быстро. Однозначно рекомендую.
@dmitry_k Kwork · Telegram Bot
★★★★★ 5.0 / 5.0
Заказывал REST API на FastAPI с авторизацией и интеграцией со сторонним сервисом. Всё сделано профессионально — покрытие тестами, CI/CD настроен. Буду обращаться ещё.
@maxim_web Kwork · FastAPI Backend
★★★★★ 5.0 / 5.0
Нужен был парсер + автоматизация с отправкой в базу. Сделал быстро, надёжно, с обработкой ошибок и логами. Не просто выполнил задачу, но и предложил оптимизации.
@anna_startup Kwork · Automation Script
★★★★★ 5.0 / 5.0
Доработка существующего Django-проекта — добавил WebSocket-уведомления и Redis-кэширование. Разобрался в чужом коде за пару часов, правки минимальные. Топ исполнитель.
@roman_tech Kwork · Django / Redis