• Файл

Heorhii

AI engineer

Розглядає посади:
AI engineer, Fullstack-розробник
Вік:
19 років
Місто:
Коломия

Контактна інформація

Шукач вказав: ТелефонЕл. поштуМесенджер

Прізвище, контакти та світлина доступні тільки для зареєстрованих роботодавців. Щоб отримати доступ до особистих даних кандидатів, увійдіть як роботодавець або зареєструйтеся.

Завантажений файл

Версія для швидкого перегляду

Це резюме розміщено у вигляді файлу. Ця версія для швидкого перегляду може бути гіршою за оригінал резюме.

Heorhii Priadkin
Fullstack Developer | AI/LLM Engineer
Kolomyia, Ukraine | Remote
[відкрити контакти](див. вище в блоці «контактна інформація») | priadk.in | GitHub | LinkedIn | Telegram

PROFESSIONAL SUMMARY
Full-Stack Developer and AI/LLM Engineer with 2+ years of commercial experience shipping production SaaS platforms
and, in parallel, designing and building LLM-powered systems end to end — architecture, backend, frontend, infrastructure,
and the LLM layer itself. Built a multi-agent legal automation platform (MCP server, 32-tool orchestrator, custom OpenAI-
compatible LLM proxy with provider routing and PII anonymization) and an omnichannel AI customer support platform
(multi-channel dialog engine with anti-hallucination guardrails, autonomous email support, compliant bulk messaging) —
both live in production on real client traffic. Comfortable across the full LLM engineering stack: function calling and tool
use, RAG and full-text search, prompt engineering, vision-OCR, agent orchestration, and cost/latency-driven provider
selection. Core backend skills in Python (FastAPI, async SQLAlchemy, Celery/arq), PHP (Laravel), and Node.js, with
React/Next.js and React Native on the frontend. Daily driver of AI coding agents (Claude Code, OpenAI Codex, Cursor,
GitHub Copilot) to accelerate delivery while maintaining code quality through rigorous review.

TECHNICAL SKILLS
AI / LLM Engineering: Multi-provider LLM orchestration and routing, function calling / tool use, MCP (Model Context
Protocol) server and agent design, agent orchestration, RAG, full-text search, prompt engineering, anti-hallucination
guardrails, vision-OCR, custom LLM proxy design (streaming, PII anonymization, cost tracking, failover), provider/model
benchmarking
Programming Languages: Python, PHP, JavaScript, TypeScript, Java, Kotlin, C#, SQL, HTML5, CSS3
Backend: FastAPI, Django, async SQLAlchemy 2.0, Alembic, arq, Celery, Laravel 11, Node.js, Express.js, ASP.NET
Core, Spring Boot, RESTful APIs, WebSockets, SSE, Queue Workers, Background Jobs
Frontend: React, Next.js (App Router, Server Components, Server Actions), React Native, TypeScript, Tailwind CSS,
shadcn/ui, Framer Motion, Blade, jQuery, Vite, Responsive Design, SPA, SSR
Databases: PostgreSQL, MySQL, MariaDB, Redis, SQLAlchemy ORM, Django ORM, Eloquent ORM, Database Design,
Query Optimization, Migrations
DevOps and Tooling: Docker, Docker Compose, Caddy, Git, GitHub, Nginx, Linux VPS, CI/CD, npm, pip, Composer,
Version Control
AI Coding Agents and Developer Tools: Claude Code (Anthropic CLI agent), OpenAI Codex (CLI agent), Cursor IDE,
GitHub Copilot, Gemini CLI, Aider, agentic workflows, multi-file refactoring, codebase-aware development, AI-assisted
code review, CLAUDE.md / AGENTS.md context design
Third-Party Integrations: Stripe, PayPal, Razorpay, Mollie, YooKassa, Flutterwave, Paystack, Authorize.Net, Twilio
SMS, Zoom API, Telegram Bot API (aiogram), ManyChat, WooCommerce API, Microsoft Graph, OneDrive (rclone),
Google Calendar, DeepL API, Tavily, Odorik SMS
Architecture and Methodologies: Multi-tenant SaaS Architecture, Multi-agent Systems, Event-Driven Architecture, Role-
Based Access Control (RBAC), Observer Pattern, State Machines, Double-Entry Accounting, Queue-Based Async
Processing, Audit Logging, Modular Architecture, Agile, Scrum, Code Review, Unit/Integration Testing

SELECTED AI/LLM PROJECTS
Omnichannel AI Customer Support Platform (E-commerce)
Independent client project | Remote 2025 - Present
● Built an AI support and sales system covering chat, email, and social channels for an e-commerce store, with one
shared dialog core and channel adapters (Telegram/aiogram, web widget, Instagram/Facebook via ManyChat) so
catalog, pricing, and escalation logic is never duplicated.
● Designed a dialog engine with three independent response checks before send: groundcheck (blocks unsupported
claims — facts must trace to real data), langcheck (response language), and promisecheck (the bot never commits
to actions it can't perform) — hallucination control enforced in code, not prompt wording.

● Built an autonomous email support bot (IMAP/SMTP) with subject classification across 8 categories, language
detection with signature localization across 11 languages, correct thread handling (In-Reply-To/References), and
three operating modes from draft-only to fully autonomous sending.
● Implemented compliant bulk email with a DB-backed queue, dual sliding-window rate limits to stay under hosting
quotas, HMAC-signed unsubscribe links with List-Unsubscribe / one-click per RFC 8058, and automatic receipt
parsing from Ravelry/PayPal to keep the customer database current.
● Built an operator admin panel (Next.js, Server Components/Actions) with realtime updates over SSE, live
conversation handoff, promotions/broadcast management, and a prompt editor.
● System serves a live store with 5,963 orders and 3,428 customers across 11 languages; covered by 919 automated
tests, each tied to a real production incident so regressions don't come back; bulk sends reach 3,400+ addresses
under a strict hosting quota without domain reputation issues.
Technologies: Python 3.12, FastAPI, arq, SQLAlchemy 2.0 (async), Alembic, PostgreSQL, Redis, MiniMax/Gemini/Groq via a unified
provider interface (12 tools, function calling), aiogram, WooCommerce REST API, Tavily, Next.js 15, TypeScript, Tailwind, shadcn/ui,
Docker Compose, Caddy.

MACEK.LEGAL — AI Document Automation Platform for a Czech Law Firm
Independent client project | Remote 2025 - Present
● Built a legal-tech platform end to end — architecture, backend, frontend, infrastructure, and the LLM layer —
automating document generation and contract review for a Czech law firm.
● Shipped 18 document types across 8 generation modules (price quotes, pre-litigation claims, NDAs, contracts,
powers of attorney, real-estate documents, court filings) and a contract review tool that flags risks and missing
clauses in incoming third-party contracts, with anchored inline edits and DOCX export using real tracked changes
(w:ins/w:del).
● Designed an MCP (Model Context Protocol) server exposing 32 tools to an in-house agent orchestrator ("Hermes")
that drafts and edits documents, tracks case deadlines, searches firm knowledge and case law, and reads scanned
documents through a vision model.
● Built a custom OpenAI-compatible LLM proxy as a control point: PII anonymization before any data leaves the
network, provider routing (MiniMax/OpenRouter) with per-model hoster pinning after finding 23x speed variance
and inconsistent output quality across hosters, streaming, dead-connection detection, and per-document cost
tracking.
● Synced a searchable case-law database of 16,814 court decisions from justice.cz with full-text search that accounts
for Czech morphology and diacritics-insensitive matching.
● Built a firm-knowledge feedback loop: a lawyer flags an error or dictates a rule, and it's applied to all future
documents via a two-stage pipeline (deterministic lookup first, sub-agent only on a match) to keep correction cost
and drift low.
● Contract generation runs in 98 seconds versus an earlier 628s baseline, and contract review in 39 seconds versus
695s, achieved through provider/hoster selection and reasoning-effort tuning rather than cutting limits; cost per
document is roughly one cent; the flagship contract type covers 20 of 20 required clauses in self-review.
Technologies: Python, FastAPI, SQLAlchemy (async), PostgreSQL, arq, Redis, httpx, Pydantic, React, TypeScript, Vite, i18n (cs/en),
python-docx, PyMuPDF, vision-OCR, Microsoft Graph, OneDrive (rclone), ARES, justice.cz, Tavily, Docker Compose, Caddy.

WORK EXPERIENCE
Fullstack Software Developer
LamaPixel (Aspector) | Remote November 2025 - Present
● Designed and shipped 18+ business modules for Jobsi, a multi-tenant SaaS ERP/CRM platform for staffing
agencies, including Human Resource Management (HRM), Customer Relationship Management (CRM) with
Kanban pipelines, Finance with double-entry accounting, hierarchical Cashbox, Projects with time tracking, Fleet
Management with GPS tracking, Recruitment with form builder, Point of Sale (POS) with warehouse management,
and Document Generator.
● Integrated 30+ payment gateways including Stripe, PayPal, Razorpay, Mollie, YooKassa, Flutterwave, and
Paystack to support global SaaS billing across multiple regions.
● Implemented an AI-powered OCR pipeline using Google Gemini API to batch-process passports, receipts, and
vehicle documents (up to 20 items per request), significantly reducing manual data entry for end users.

● Developed a token-based billing system backed by Stripe for metered AI-scan consumption, including token
balance, transactions, orders, and configurable pricing.
● Built a dedicated mobile-adaptive UI layer with Blade templates and Tailwind CSS without an SPA framework,
featuring bottom-sheet modals and collapsible sections for one-handed usability.
● Localized the platform into 5 languages (Russian, English, Ukrainian, Polish, Czech) using JSON and PHP
dictionaries.
● Designed Infinity SPA Ecosystem: a production-grade 3-role CRM (Admin, Operator, Masseuse) with isolated
RBAC permissions, synchronized bidirectionally via REST API with a public WordPress booking website for a
Czech market client.
● Implemented an end-to-end SMS verification flow via Odorik API with an async retry queue (3 attempts), phone-
number normalizer, low-balance Telegram alerts, and full audit logging.
● Engineered state-machine order flow, operator shift system with real-time conflict resolution for overlapping shifts,
per-operator bonus rates, and localized Telegram notifications in Czech.
Technologies: PHP 8.2, Laravel 11, MySQL, React Native, Blade, Tailwind CSS, WordPress, Redis, Google Gemini API, Stripe API,
Twilio, Zoom API, Google Calendar, Odorik SMS, Polylang, Spatie Laravel Permission.

Freelance Software Developer
Self-employed | Remote February 2025 - November 2025
● Delivered 11+ cross-platform projects across 4 environments (Android, Windows desktop, Minecraft Java servers,
web) using 5+ programming languages, focusing on complex technical challenges that required deep expertise.
● Developed AntiCorruption, a real-time audio, video, and GPS streaming system using Kotlin for Android and the
RTMP protocol, paired with a C# WPF desktop admin dashboard featuring live map tracking and auto-recording.
Stack: Kotlin, C#, WPF, MariaDB.
● Built BusinessGladiolus, a commercial Java plugin for Spigot/Paper Minecraft servers implementing a full
economic simulation with player businesses, WorldGuard-protected territories, banking with loans and bankruptcy
confiscation, anti-monopoly rules, and government auctions with dynamic pricing. Stack: Java, Spigot/Paper,
WorldGuard, Vault.
● Developed TranslateGladiolus, a commercial Java plugin for real-time Minecraft chat translation via DeepL API
supporting 11+ languages, with async processing, MySQL-backed 7-day translation cache via HikariCP, and 8
granular access levels. Stack: Java, Spigot/Paper, DeepL API, MySQL, HikariCP.
● Engineered CraftNova Launcher: a custom Minecraft launcher built from scratch using Electron, Java, and Node.js,
featuring integrated anti-cheat, server access control, auto-updates, and a modern UI.

EDUCATION
Bachelor of Software Engineering
Dnipro University of Technology, Ukraine | 2024 - 2028 (expected)

LANGUAGES
Ukrainian (Native), Russian (Native), English (B1+)

Інші резюме цього кандидата

Схожі кандидати

Усі схожі кандидати

Кандидати у категорії

Кандидати за містами


Порівняйте свої вимоги та зарплату з вакансіями інших підприємств: