Skip to content
Nazar Pauk
Menu
Українською

Ukraine · Remote

Nazar Pauk

AI Implementation Engineer

I connect LLMs, CRMs, databases, and messengers into working business processes.

About

I implement AI agents, MCP tools, and automation pipelines for enterprise clients and e-commerce teams. My work sits between product operations and engineering: I map a messy process, then ship the integration that removes the manual steps. A robotics degree keeps the systems thinking; the day-to-day is JavaScript, APIs, RAG, and the tools operators actually use.

Skills

AI & LLM

  • AI integration
  • AI Agents
  • RAG
  • LLM-based Systems
  • Prompt Engineering
  • AI Workflow Design
  • Context & Memory Management
  • Generative AI
  • MCP
  • Langfuse
  • Dify
  • VoltAgent

Automation & Integration

  • Make.com
  • n8n
  • APIs
  • System Integration
  • JSON
  • Telegram Bot API
  • KeyCRM API
  • OpenCart REST

Data & BI

  • MySQL
  • Databases
  • Metabase
  • Grafana
  • DBeaver

Development

  • JavaScript
  • PHP
  • Google Apps Script
  • HTML/CSS
  • React
  • C++
  • Docker
  • Git/GitHub
  • Cursor AI

Process

  • Jira/Confluence
  • ClickUp
  • QA

Experience

AI Implementation Engineer

ConnectiveOne · Full-time

Dec 2025 — Present
Kyiv, Ukraine · Remote

  • Engineered and maintained automation scenarios, AI agents, and Telegram bots for large-scale enterprise clients (Megogo, Glo, Velo).
  • Implemented the Model Context Protocol (MCP) to analyze scenarios and build custom tools for the Instance Agent, automating platform actions.
  • Built internal productivity tools with Cursor AI, shipped Dify-based AI solutions, and integrated VoltAgent (RAG) for contextual retrieval.
  • Designed custom integration modules from scratch, including authentication and account-binding workflows for Telegram bots.
  • Administered MySQL databases and configured Metabase dashboards to monitor and optimize internal processes.
  • Refactored legacy automation scenarios and created specialized AI tools to accelerate QA against strict rules.

Stack

  • JavaScript
  • MCP
  • Dify
  • RAG
  • Langfuse
  • Metabase
  • Grafana
  • MySQL
  • Docker
  • Cursor AI

Business process automator

Dolota LLC · Part-time

Jun 2025 — Present
Boryslav, Lviv, Ukraine · Remote

  • Automated the full product lifecycle (KeyCRM → AI-generated descriptions → OpenCart publishing) with zero manual data-entry steps.
  • Integrated Make.com and the OpenAI GPT API for bilingual (UA/EN) SEO content with manager-driven prompt adjustments.
  • Built a KPI engine with a configurable bonus matrix, auto-archiving, and automated Telegram reporting for the sales team.
  • Developed a PHP webhook controller for OpenCart with multi-language product descriptions and automated image-gallery processing.
  • Created a delivery-readiness checker that validates product dimensions, photos, and purchase prices before shipment.

Stack

  • Make.com
  • n8n
  • OpenAI API
  • KeyCRM API
  • OpenCart REST
  • PHP
  • Google Apps Script
  • MySQL
  • Telegram Bot API

Case studies

E-commerce product pipeline

Problem
Catalog updates required managers to copy product data between KeyCRM and OpenCart by hand — slow, error-prone, and bilingual SEO never kept up.
Solution
Make.com + OpenAI generate UA/EN descriptions from CRM fields; a PHP webhook publishes products and galleries to OpenCart. Managers only adjust prompts when the tone needs a tweak.
Impact
Zero manual data-entry steps across the product lifecycle.
  • Make.com
  • OpenAI API
  • KeyCRM API
  • OpenCart REST
  • PHP
  • MySQL

Sales KPI and bonus engine

Problem
Sales bonuses lived in spreadsheets. Rules drifted, archives were missing, and the team learned results too late.
Solution
A configurable bonus matrix with auto-archiving and a Telegram bot that reports KPIs to the sales team on a schedule.
Impact
One source of truth for bonuses and automated reporting instead of weekly spreadsheet archaeology.
  • Google Apps Script
  • Telegram Bot API
  • MySQL

MCP tools for the Instance Agent

Problem
The platform agent could talk, but it could not reliably take the actions operators needed on each client instance.
Solution
Custom MCP tools that inspect automation scenarios and expose instance-specific actions the agent can call instead of ad-hoc prompts.
Impact
Platform actions became first-class tools — inspectable, repeatable, and tied to real scenarios.
  • MCP
  • JavaScript
  • Cursor AI
  • Dify

Enterprise Telegram bots

Problem
Large clients needed bots that were not chat toys: they had to authenticate users, bind accounts, and trigger real platform workflows.
Solution
Custom integration modules with auth and account-binding, plus AI agents on top of existing automation scenarios for Megogo, Glo, and Velo.
Impact
Operators get a messenger interface to production systems without handing out admin panels.
  • Telegram Bot API
  • AI Agents
  • JavaScript
  • MCP

Delivery readiness checker

Problem
Shipments left the warehouse with missing photos, wrong dimensions, or no purchase price — and logistics found out too late.
Solution
An automated checker that validates dimensions, photos, and purchase prices against the catalog before a shipment is allowed to proceed.
Impact
Incomplete products are blocked before they become a courier problem.
  • Make.com
  • OpenCart REST
  • KeyCRM API

Education & courses

  1. 2023 — 2027

    LNUVMB named after S.Z. Gzhytskyi

    Bachelor’s in Automation of Computer-Integrated Technologies and Robotics

  2. 2024 — 2026

    Scrimba

    Learn JavaScript, Learn React, Introduction to ES6+, Learn SQL

  3. 2025

    Vadym AI

    Automate routine tasks with AI

  4. 2022

    Sigma Software University

    Interesting Front-end

Contact