Contact information

You can get this candidate's contact information from https://www.work.ua/resumes/17680326/

Work experience

Middle Python Developer | AI / LLM

from 04.2026 to now (6 months)
Undertalk, кол-центр, Дистанційно (Телекомунікації, зв'язок)

Building a multi-module call-center CRM from scratch — independent modules over a shared operator/client data layer, in a microservice layout with Celery / RabbitMQ / Redis workers.

Delivered end-to-end in the last two months: operator workspace, discipline control, telephony analytics, shift scheduling and recruiting. I own the full cycle on each — requirements with the client, spec, database schema, implementation, tests, production deployment and incidents. Vue.js frontend where a module needs it.

- Shipped a production MCP server — 38 tools that let Claude configure payroll rules from chat: schema introspection, validation, salary simulation on real data, gated write path. MCP runs as a second transport over the same executors as the API.

- Hardened it for production: the write tool refuses to persist unless validation and a money-impact simulation both passed — any edit resets the gate, so an agent can't save a config no human priced. OAuth 2.1 (Auth0), connector forwarding the caller's own token.

- Built the in-CRM AI copilot on the same tools — an Anthropic tool-use loop that self-corrects on tool errors.

- Built operator analytics and client analytics as separate subsystems, computed per connected CRM and aggregated across all of them. Client complexity is a weighted coefficient over three normalized metrics, recomputed nightly.

- Integrated Binotel telephony, external CRMs and order sources, and Viber chat (text, media, MinIO) — via webhooks, polling APIs and queues, with idempotent syncs and an identity-mapping layer that resolves the same operator or client across systems that identify them differently.

- Designed a configurable discipline rules engine (three parameterized rule types — new thresholds ship as configuration, not code) and the QA subsystem behind it: pause accounting, a violation state machine and an audit trail.

Stack: Python · FastAPI · PostgreSQL · SQLAlchemy async · Celery · RabbitMQ · Redis · Claude API · MCP · Docker · Vue.js

Python developer

from 01.2025 to 03.2026 (1 year 3 months)
SECL Group, Дистанційно (IT)

Backend developer (Python) on a production web platform, working inside a cross-functional team under a strict PR-review process.



Worked across several backend modules — authentication, client records, notifications and reporting — implementing endpoints, data access and background logic, and taking each one through review to release.



- Shipped Alembic migrations against a live production database with zero downtime.



- Profiled and rewrote the slowest queries on the highest-traffic endpoints.



- Replaced browser-side polling of our own API with a WebSocket layer for real-time push notifications, and added Redis caching on hot endpoints — cut redundant requests and reduced server load.



- Integrated third-party services over REST APIs, including Google Calendar (OAuth 2.0, event synchronisation).



- Tracked down and fixed recurring production incidents in JWT auth and async session handling — session lifecycle, transaction boundaries and data consistency under concurrent load.



- Every change reviewed and tracked in Jira; ran Docker-based production deployments on Ubuntu.



Stack: Python · FastAPI · PostgreSQL · SQLAlchemy async · Alembic · Redis · WebSockets · Pydantic V2 · Docker · Linux

Education

Online School, ProgAcademy

IT, Київ
Higher, from 2023 to 2023 (less than 1 year)

Python Backend Developer

Additional education and certificates

No additional education or certificates.

Knowledge and skills

• Здатність до навчання
• Python
• Git
• SQL
• Знання принципів ООП
• PostgreSQL
• Docker
• SQLite
• REST
• FastAPI
• GitHub
• Alembic
• Asyncio
• SQLAlchemy
• Unit Of Work
• Pydantic
• Clean Architecture
• REST API
• Django
• CSS
• HTML
• MySQL
• Комунікабельність
• Redis
• JSON Web Token
• Docker Compose
• Asynchronous programming
• Pandas
• Celery
• WebSocket
• JavaScript
• Node.js
• RabbitMQ
• Postman
• Google API

Language proficiencies

• English — average
• German — beginner
• Ukrainian — fluent
• Polish — fluent

Additional information

Python backend developer with 2 years of commercial experience building production
async systems on FastAPI — and shipping LLM features that run in production, not in demos.

Right now I'm building a multi-module call-center CRM from scratch at UnderTalk:
microservices over a shared operator/client data layer, Celery / RabbitMQ / Redis workers,
Docker and Docker Compose across dev and prod. In the last two months I delivered five
modules end-to-end — operator workspace, discipline control, telephony analytics, shift
scheduling and recruiting — each from requirements with the client through schema design,
implementation, tests and production deployment. Plus the analytics subsystem behind the
operator and client dashboards.

The part I'm proudest of: an MCP (Model Context Protocol) server with 38 tools that lets
Claude configure live payroll rules from chat. OAuth 2.1 with Auth0, a write path that
refuses to save until validation and a money-impact simulation both pass, concurrency-safe
per-session state. The same tools sit behind an in-CRM AI copilot — an Anthropic tool-use
loop that self-corrects on tool errors.

Integrations are a large part of what I do: telephony (Binotel), an external CRM and order
source, Viber messaging — over webhooks, polling APIs and message queues, with idempotent
syncs so a retried event never produces duplicates.

Side projects, all with real test coverage and CI: Trend Radar (Playwright scraping +
5 LLM providers with deterministic fallback, 231 tests), a full e-commerce platform live
on AWS EC2, a RAG pipeline over PDFs, and a Telegram job bot.

Code: github.com/bogdan0089

Open to Python Backend / AI Engineer roles — remote.

Ukrainian (native) · Polish (C2) · English (B2)