Taras

CONTACTS Taras Kozak
Wrocław, Poland AI Automator | n8n & AI Agents
[відкрити контакти](див. вище в блоці «контактна інформація»)
[відкрити контакти](див. вище в блоці «контактна інформація») Junior AI Automation Engineer with hands-on experience designing AI agents and end-to-
Telegram end business automation using n8n, LLMs, and no-code integrations. Built 4 production-
Portfolio (Notion) grade systems covering voice AI, call analytics, lead processing, and support-desk
automation.
HARD SKILLS Experienced with OpenAI and Anthropic APIs, AssemblyAI, webhooks, Zoho CRM, Klaviyo,
and Google services, with a strong focus on prompt engineering, structured outputs,
• Prompt Engineering (multi-block
prompts, chain-of-thought,
JavaScript data transformation, and measurable business value.
structured outputs / JSON Schema)
• AI Agent & Workflow Architecture PROJECT EXPERIENCE
(multi-workflow systems, function
calling)
Smart Call Auditor — AI Call Analytics & Conversational
Tools: n8n, AssemblyAI, OpenAI, JavaScript, Google Sheets API, Notion API
• No-Code / Low-Code Automation
(n8n: triggers, IF/Switch, error • Built a 3-workflow autonomous ETL pipeline converting raw call recordings into
handling) structured analytics (STT → AI classification → reporting).
• API Integration (REST, webhooks, • Implemented AssemblyAI Universal-2 transcription for Ukrainian with smart polling
OAuth, polling logic) (Wait + IF) for long calls.
• Data Transformation & Reporting • Designed an AI classifier segmenting calls by tone and topic; every call fully evaluated
(JavaScript, arrays, metrics) in under 5 minutes.
• Speech-to-Text pipelines & Voice AI • Developed a conversational BI agent with function calling (get_call_stats) that
agents
computes date ranges from natural language and returns ready reports in chat.
• Business Process Analysis (5 Whys, Result: fully automated call analytics with zero manual listening; saves the company
ROI, automation readiness)
$700/month, project payback in 3 months.
TOOLS AI-Driven Lead Processing System — Barbershop
• n8n Tools: n8n, OpenAI / LangChain, Zoho CRM, Klaviyo, Google Sheets, Telegram, JavaScript
• OpenAI API • Built a webhook-based intake pipeline with custom JavaScript parsing of raw form
• Anthropic Claude API submissions.
• LangChain nodes, AI Agents • Implemented AI lead scoring (hot / warm / cold) from free-text customer comments
with self-correcting fallback logic.
• AssemblyAI (STT)
• Synchronized each lead simultaneously to Google Sheets, Zoho CRM, and Telegram
• Happ.tools (Voice AI)
(HTML-formatted notifications).
• Zoho CRM, Klaviyo
• Automated Klaviyo segmentation by lead status to trigger targeted email flows (promo
• Google Sheets / Drive API codes for hot leads, nurture content for cold).
• Notion API Result: replaced 10–15 minutes of manual work per lead; zero data loss across three
• Telegram Bot API, Slack API, Gmail systems; prioritized call queue for sales.
API
• Supabase (PostgreSQL) AI Ticket Routing & SLA Escalation — E-commerce Support Desk
• JavaScript, JSON Schema Tools: n8n, OpenAI (GPT-4.1-mini), Structured Output Parser, Google Sheets, Slack, Gmail
• Designed a 5-block classifier prompt assigning category, priority (P1–P4), owner, SLA,
SOFT SKILLS and a written justification for every ticket, scored across urgency / business impact /
customer context.
• Analytical
• Enforced clean data with JSON Schema validation (enum + required constraints on 7
• Detail-oriented
fields).
• Self-driven
• Built a scheduled SLA watcher with a 3-level escalation chain: owner reminder → team
• Systematic problem-solver lead → director email with full context.
• Verified logic with 10 end-to-end test scenarios, including VIP cancellations and
LANGUAGES
payment failures.
• English — B2 Result: eliminated manual triage for ~120 tickets/day; SLA breaches escalate within
• Ukrainian — Native minutes instead of days; scales without added headcount.
• Polish — B2 AI Voice Sales Qualification System — Electronics Retailer
Tools: Happ.tools (Voice AI), n8n, Anthropic Claude, Zoho CRM, Webhooks
• Designed a multi-workflow architecture (init, in-call tools, post-call processing) for an
AI voice agent qualifying inbound sales leads.
• Implemented automatic CRM field population from live conversation data ($fromAI
mapping to Zoho CRM).
• Built post-call payload processing: transcript storage, structured lead-data extraction,
 and handoff to sales.
Result: automated first-touch lead qualification by phone — structured, CRM-ready lead
data without manager involvement.

EDUCATION
GoIT — AI Automation Program
2025 – 2026
Lviv Polytechnic National University — Master’s Degree in Computer Engineering
2007 – 2012