Bohdan

AI AUTOMATION & APPLIED AI ENGINEER
Email: [открыть контакты](см. выше в блоке «контактная информация») Mobile: [открыть контакты](см. выше в блоке «контактная информация») Telegram: @Pah0m4k Remote — Portugal (EU)
Profile
Automation engineer who owns the full cycle: turning vague business asks into feasible scope, architecture, and shipped systems
— usually as the only technical person on the project.
Pragmatic by default: I assess what is actually buildable, propose the version that delivers most of the value at a fraction of the
cost, and reserve LLMs for tasks where deterministic logic cannot win — keeping systems predictable and cheap to run.
Recent impact: replaced a licensed SaaS helpdesk with an in-house support platform (est. ~$50K/year in licensing removed), and
shipped AI agents that took recurring finance, legal, and support work off people's hands.
Experience
AI Automation & Applied AI Engineer Jul. 2024 – Jun. 2026
Fintech company (name under NDA) — Remote
– Owned projects end-to-end. Received business-level briefs with no technical grounding, ran the feasibility assessment, and
returned a realistic plan: scope, architecture, stack, and phasing. One request would have taken ~3 years as specified; I
delivered a working alternative that covered the actual business need — saving the company quarters of wasted engineering.
– Cut infrastructure cost by design. Chose deterministic logic, rules, and classic integrations wherever they beat an LLM,
reserving AI for tasks that genuinely required it — lowering ongoing inference cost and eliminating a class of non-
deterministic failures.
– Shipped to production, not to prototypes. Used n8n / Make / Flowise to validate ideas fast, then migrated proven solutions
into Python (FastAPI) backend services with async processing, webhook handling, and proper error recovery.
– Freed up entire departments. Built integration-driven automations across CRM, databases, messengers, email, and
document platforms for product, legal, finance, and operations — removing recurring manual work from each of them.
– Raised the reliability bar. Made validation, retries, error handling, and monitoring standard across automations, so they run
unattended on business-critical data.
– Embedded in the analytics team. Built and maintained a large set of recurring reports and automated alerting, giving the
business continuous visibility into its key metrics instead of ad-hoc manual pulls, and flagging anomalies the moment they
appeared rather than at month-end.
– Volume of small wins. Delivered dozens of micro-automations on top of the flagship projects — each one small, each one
removing a recurring manual step; together they add up to a meaningful amount of time returned to the teams every week.
AI Automation Specialist — N5Bank Jun. 2023 – Jun. 2024
– Built AI-driven pipelines connecting content generation, document workflows, and voice interaction into unified systems used
by the business daily.
– Automated contract handling via DocuSign (Google Forms → dynamic template filling → e-signature → client notification →
storage), removing manual document preparation and the errors that came with it.
– Delivered a custom AI voice assistant (ElevenLabs, voice cloning, contextual responses) integrated with backend modules for
natural client communication.
– Automated LinkedIn and Instagram publishing end-to-end (Gemini API: visual + caption + scheduling), turning a recurring
manual routine into a hands-off process.
– Worked directly with DevOps and Backend teams on API orchestration, webhooks, and data consistency — translating
between business intent and technical design.
– Implemented dozens of smaller automations across day-to-day operations — individually simple, collectively removing a
steady stream of manual work from the team.
Freelance AI Automation Developer Jun. 2022 – Jun. 2023
– Delivered full-cycle automations for 6 companies of different sizes — from brief to production release and ongoing support.
– Launched 40+ automation solutions and 30+ documented workflows that clients actually kept running, rather than one-off
scripts.
Selected Projects
In-house Support Team CRM (helpdesk platform) Sole engineer
– Business case: The company was paying recurring per-seat licensing for a third-party helpdesk (est. ~$50K/year) with no
data ownership and no way to build its own AI layer on top. I designed and built the in-house replacement end-to-end —
architecture, stack, security model, product scope, and implementation — with a project manager involved only in gathering
partner-integration requirements.
– Stack: NestJS (TypeScript, strict) + Vue 3 / Vite / Tailwind, Prisma, self-hosted Supabase (Postgres + Auth), Redis + BullMQ,
Caddy (auto-HTTPS), Docker, GitHub Actions CI.
– Built for scale and safety: multi-tenant isolation enforced at the ORM layer, RBAC with MFA, audit logging, and list
virtualisation for a 370K+ ticket dataset; full migration of historical tickets, macros, automations, routing rules, SLAs, and
taxonomy from the legacy platform, verified on staging before production.
 – Security as a release gate: ran a structured security review (39 findings, categorised by severity) and blocked any real
customer data until every critical item was closed — PII scrubbed from logs, encrypted off-host backups with tested restore,
egress allow-list, non-root containers.
– Design system: token-driven, white-label-ready UI (Vue 3, shadcn-vue, Reka UI, Tailwind) so the platform could carry
multiple brands without a UI rewrite.
AI Support Widget (customer-facing)
– LLM assistant embedded on the company website, answering customer questions about the platform, payments, and account
topics — deflecting routine support tickets before they reach an agent, and feeding directly into the CRM as a ticket channel.
Finance AI Agent
– Agent for the finance department handling ~200 invoices/month: invoice generation, document handling, data aggregation,
and period-based summaries — turning a recurring manual cycle into a supervised, automated one.
Legal Document Assistant (LLM)
– LLM-based assistant for legal managers processing ~50 documents/month: analysis, structuring, and risk detection — giving
lawyers a pre-read instead of a blank page and shortening review time per document.
Geospatial Store-Locator Automation
– System for a nationwide cosmetics distribution network that accepted location input at any level of precision (coordinates,
street, district, metro station, landmark), normalised it via custom scripts, and used Google Maps APIs (geocoding, places,
distance matrix, directions) to return the nearest store, an N-nearest list, or a route — removing a manual lookup step from
customer-facing flows.
R&D / Prototypes
– AI data agent over AWS Athena. Designed and validated an approach for LLM-driven analytics under hard constraints: no
native schema introspection, async-only query execution, and a security policy forbidding data replication into an
intermediary store. Built a context-retrieval and execution layer, proven on a subset of tables; rollout deprioritised in favour of
higher-ROI work. The same approach was later productised by AWS in Bedrock Agents.
– AI landing-page generator. Prototype where an LLM analysed competitor pages and adapted them to our brand;
requirements were captured conversationally through a Slack agent, and the generated page could be finished by hand in a
lightweight SPA editor. Shelved after the scoping phase in favour of higher-priority work.
– Text-to-speech pipeline. Built a TTS integration for automated voice output within internal workflows.
Skills & Tools
– Backend & Programming — Python (FastAPI, Flask, AsyncIO, OOP, REST, webhooks); TypeScript / NestJS; JavaScript; API-
first architecture; async & background processing (Redis, BullMQ); authentication and service integration.
– Automation & Orchestration — n8n (incl. custom nodes and custom logic), Make, Flowise; event-driven and webhook-
based automation; reusable templates, validation layers, retries, error handling, and monitoring.
– AI & LLM Engineering — LangChain, LangGraph (routed, validated, structured-output pipelines); OpenAI & Gemini APIs;
RAG and context-retrieval design; prompt engineering; OCR pipelines; LLM-based classification, extraction, and
summarisation; cost-aware AI design.
– Machine Learning — PyTorch, scikit-learn, NumPy, Pandas; dataset preparation and preprocessing; fine-tuning; supervised
learning pipelines; model evaluation and inference integration.
– Data & Infrastructure — PostgreSQL, MySQL, Supabase, AWS Athena; Prisma; Docker; GitHub Actions (CI/CD); Caddy;
multi-tenant data isolation, RBAC/MFA, audit logging, encrypted backups.
– Integrations — CRM & helpdesk platforms (HubSpot, Pipedrive, Salesforce); Google Cloud & Maps APIs; Meta Graph,
LinkedIn, Instagram, Telegram Bot, WhatsApp Business, Slack; Stripe, PayPal; DocuSign, Google Drive/Sheets, Notion, Airtable;
Gmail, SendGrid, Mailchimp; web scraping (REST, HTML parsing, pagination, rate-limit-aware crawlers).
– Frontend — Vue 3, Vite, Tailwind, shadcn-vue / Reka UI, Pinia; token-driven design systems and white-label theming.
Education & Certifications
National Technical University of Ukraine — “Igor Sikorsky Kyiv Polytechnic Institute”
– BSc in Electrical Energy — “Renewable Energy Sources”
– MSc in Management — “Management of International Business”
Certifications
– Machine Learning Specialization — DeepLearning.AI & Stanford University (Advanced Learning Algorithms; Unsupervised
Learning, Recommenders & Reinforcement Learning)
– Fine-Tuning and Large Language Models — Hugging Face
– Java Backend Development — Prog.Academy
– n8n: AI Agents, Automation & Voice Systems
– Applied AI Research & Collaboration — AI Community Program