Денис

Denys Shcherbatyi
AI Engineer | LLM Agents · Python · FastAPI · MCP
St. Pölten, Austria · Remote (CET) | [відкрити контакти](див. вище в блоці «контактна інформація»)
Telegram: @Zebaro24 | [відкрити контакти](див. вище в блоці «контактна інформація») | github.com/Zebaro24 | zebaro.dev

SUMMARY
AI engineer with 2+ years in Python, building LLM products. Over the last year I built a production AI SaaS from
the first commit: a multi-agent LLM pipeline with validated structured outputs, fault tolerance for non-deterministic
results and an MCP server with OAuth for AI assistants. I also build with AI: Claude Code and other AI tools are
part of my everyday toolkit.

TECHNICAL SKILLS
AI & LLM: multi-agent LLM pipelines, structured outputs, MCP, prompt engineering, LLM APIs
AI-Native Workflow: Claude Code (custom subagents, skills, hooks)
Backend & Data: Python, FastAPI, Pydantic, asyncio, REST APIs, MongoDB, PostgreSQL
DevOps: Docker, GitHub Actions CI/CD, Google Cloud
Frontend: TypeScript, Next.js, React

WORK EXPERIENCE

Fullstack Engineer (Python / TypeScript) | Contract · US-based company (NDA)
Sep 2025 – Aug 2026 | Remote
Commercial AI-powered SaaS platform for content analysis, from the first commit to production.
• Built a multi-agent LLM pipeline: parallel agents whose structured outputs merge into one scored report
• Designed fault tolerance for non-deterministic AI output: results validated against strict schemas, incomplete
results never reach users, failed runs recover without repeating costly steps
• Delivered an MCP server with OAuth 2.0 so AI assistants can work with the product on users' behalf
• Built AI-native dev tooling: custom Claude Code subagents, skills and hooks behind one quality gate (lint,
types, security, architecture rules, coverage) shared by local runs and CI
• Optimized resource usage for heavy processing workloads, keeping the service stable under concurrent jobs
• Designed the architecture and built the platform end to end: FastAPI backend, Next.js / TypeScript frontend
• Built CI/CD with safe database migrations, subscription billing with usage quotas and an admin console
Scale: 1,300+ backend tests · 80+ documentation pages incl. 20+ ADRs · 1,100+ commits

Freelance Fullstack Developer | Self-employed
May 2024 – Sep 2025
Check-Point-SHOP — Telegram bot for ordering products with delivery
• Built customer and admin flows: catalog, orders, payment confirmation, order status, inventory, roles
Stack: Python, pyTelegramBotAPI, SQLAlchemy, PostgreSQL
Campus-Navigate-3D — Interactive 3D campus map built for a university · Live demo · Backend
• Django REST API and admin panel for locations, camera routes and 3D models; React + Three.js frontend
Stack: React, Three.js, Django, DRF, PostgreSQL, Docker

PROJECTS
Zebaro Core — personal platform on my own server: Telegram bot, FastAPI services, MCP server and site
2025 – present
• Deployed MCP server (Streamable HTTP, bearer auth) used daily with AI assistants to track applications
• Job search: Playwright scrapers for 9 job boards (1,700+ vacancies in 2 months) with tiered filtering
• FastAPI service: Jobs API with keyset pagination, webhooks for GitHub, Telegram and Monobank
Stack: Python, FastAPI, FastMCP, Playwright, MongoDB, Docker Compose, GitHub Actions, pytest (650+ tests)

EDUCATION & LANGUAGES
FH St. Pölten, Austria | Erasmus+ exchange semester, Computer Science | Feb 2025 – Jul 2025
Chernihiv Polytechnic National University | BSc in Computer Engineering (AI Systems) | 2021 – 2025
Ukrainian — Native | English — B2 (daily written work, spoken daily in Austria)