$ whoami

Eugene Shemchuk, Agentic AI Engineer

AI has arrived.

Eugene built this site with Claude in under 2 hours.

Need someone to turn your ideas into something real users rely on?

LLMs, cloud, APIs, Python. Trusted by Netflix, Nike and ABN AMRO.

Schedule 30 min call Fractal Drift Pixel Voyager

$ whoami --verbose

About

Portrait of Eugene Shemchuk

Eugene Shemchuk is a freelance senior backend/platform engineer with 11 years of experience building software and AI systems, including roles at Netflix and ABN AMRO. Based in Amsterdam, he specialises in agentic AI, backend platform engineering, and AI product development.

experience
11 years
past clients
Netflix, ABN AMRO, Nike, SportsShoes.com
based in
Amsterdam, NL

$ cat expertise.md

What I do

Eugene Shemchuk connects LLMs and AI models to real APIs, data, and business processes — turning agentic AI ideas into production systems.

Agentic AI systems

Design and build LLM-powered agents and conversational systems that plug into real APIs, data, and business logic — from an employee virtual assistant at a major bank to conversational platforms covering hundreds of topics.

Backend & platform engineering

Cloud migrations, API refactors, and platform reliability work across AWS and Azure, built to hold up under production traffic and support further AI capability.

AI product development

Take AI products from discovery to shipped feature: personas, experimentation, A/B testing, and alignment across engineering, data science, and UX.

Applied ML, NLP & computer vision

Applied machine learning across computer vision (up to 99.9% accuracy), NLP-based product search, and conversational AI, from data prep through production deployment.

$ ls ~/work

Selected work

View more on GitHub →

Netflix Proactive issue-discovery dashboards

problem
Technical issues across a global platform were often caught reactively, after becoming incidents.
approach
Built dashboards and automation to surface problems early, plus an AI/ML automation MVP to cut repetitive technical work.
outcome
Faster detection and resolution of cross-platform issues, feeding into org-wide reliability practices.

SAI Retail data platform & computer vision

problem
A retail client needed reusable data services plus AI-powered search and image recognition.
approach
Built a modular Python REST API platform, NLP-based product search, and a computer-vision pipeline end to end.
outcome
Image-recognition accuracy up to 99.9%, shipped from data prep through production deployment.

ABN AMRO Employee virtual assistant — cloud migration

problem
An employee-facing AI assistant needed to move off AWS onto Azure without breaking service for daily users.
approach
Led the technical migration and coordinated across engineering, data science, and UX while evolving the assistant's LLM-powered components.
outcome
Zero-disruption migration, plus new proactive, personalized messaging and refactored backend APIs that unlocked further AI features.
Microsoft case study ↗

ZEO Alliance Conversational AI system — $1M+ annual savings

problem
A business needed to cut support costs and lift customer satisfaction without adding headcount.
approach
Built a ChatScript/API.AI conversational system, trained on real conversation data, iterated with A/B testing.
outcome
Reported to save over $1M annually while increasing customer satisfaction by ~30%.
mackeeper.com ↗

Want to build something similar? Book a call

$ ls ~/pet-projects

Other work

Smaller builds by Eugene, outside client work.

Web experiences

  • Natal Chart ↗

    Natal chart tool that casts a birth chart from name, birthplace, date and time in a few short steps, showing planets, elements and the closest aspects. Works without a birth time by casting for noon. Leads into booking a paid one-to-one reading with Alex, the site's owner and astrologer. Designed and built by Eugene.

  • WAO — We Are One ↗

    Scroll-driven web experience for the dvaaya community that slows visitors down: a guided three-breath pause, then a single card with one small act to take into the day. Available in English, Ukrainian and Russian, and readable without JavaScript. Designed and built by Eugene.

Video automation

  • Automated captions and video editing

    Send a video, a voice note or a line of text to a Telegram bot and get back a finished vertical clip with word-by-word captions, voiceover and music. Runs fully on-device: no LLM calls, no per-clip cost.

Web graphics

  • Pixel Voyager

    Tiny 8-bit space shooter drawn on a low-res canvas with no images or libraries. Mine asteroids for crystals, level up your ship, break planets and escape black hole gravity.

  • Fractal Drift

    Live Mandelbrot zoom rendered with WebGL in the browser, diving up to 240,000× into the set. Click or scroll to steer it, and switch colour styles and warp.

Telegram bots

  • Baker Bot ↗

    Telegram bot that stores and serves Eugene's bread recipes in an easy-to-browse format.

  • Covid Tracker Bot ↗

    Telegram bot that tracked and reported COVID-19 case counts per country.

  • Universal Link Bot ↗

    Telegram bot that converts a song link into a universal link, so friends can open it on whichever music platform they use.

Instagram automations

  • dvaaya ↗

    ManyChat automation for lead generation, guiding people through grounding meditation techniques and a flow of associative cards for intuitive work.

  • siorbaty.mozhna ↗

    ManyChat automation handling lead generation for online tea ceremony and tea education workshops.

iOS & computer vision

  • Nike Shoe Recognizer ↗

    iOS app using Swift and Core ML that identifies Nike sneakers from a photo — self-collected training data, model, and app built end to end.

$ cat references.log

What people say

"One of the most exceptional Data Scientists I have encountered in my 14 years of experience... His endless curiosity, combined with his willingness to take risks, leads me to believe there will be no limit to his growth and achievements."

Juan Martin Maglione, Head of Technology and Artificial Intelligence, SAI

"Eugene has proved himself as a technical and analytical person with a fantastic collaborative spirit and drive."

Amin Gholoobi, Technical Escalation Manager, Netflix

Additional references from previous employers are available on request.

$ cat faq.md

FAQ

What does Eugene do?

Eugene is a freelance senior backend/platform engineer based in Amsterdam, specialising in agentic AI systems, backend and platform engineering, and AI product development.

Is Eugene available for freelance work?

Yes — Eugene is currently available for freelance and contract engagements.

How much experience does Eugene have?

11 years, spanning software engineering, applied ML/NLP/computer vision, and large-scale platform engineering.

What companies has Eugene worked with?

Netflix, ABN AMRO, Nike, and SportsShoes.com, spanning the entertainment, banking, and retail industries.

What kind of results has Eugene delivered?

Examples include a conversational AI system reported to save $1M+ a year and lift customer satisfaction by 30%, and computer vision systems reaching up to 99.9% accuracy.

Where is Eugene based, and does he work remotely?

Eugene works fully remote with no fixed location ties. He holds legal residence in Amsterdam, Netherlands, where he is registered as a freelancer (eenmanszaak).

What AI and LLM technologies does Eugene work with?

Agentic workflow design, Model Context Protocol (MCP) server development, Claude API and Claude Code, the OpenAI API, prompt engineering, n8n orchestration, and conversational AI/NLP.

Has Eugene built MCP-based or agentic systems?

Yes — including an MCP server exposing database operations and Claude-powered creative-generation agents as callable tools for Claude Desktop and Claude Code, published as open source on his GitHub.

What cloud platforms has Eugene worked with?

AWS, Microsoft Azure, and Google Cloud Platform — including building the entire Azure side of ABN AMRO's platform during its AWS-to-Azure migration.

Does Eugene have computer vision or mobile ML experience?

Yes — an iOS image-recognition app built with Swift and Core ML reaching 98% accuracy in shoe-model detection, and a computer-vision security system running real-time inference on an Nvidia Jetson Orin Nano.

How does Eugene evaluate AI systems?

Through A/B testing with NPS tracking, held-out validation, error analysis, and human review of low-satisfaction interactions.

What other measurable results has Eugene delivered?

At Netflix, cut incident-response SLA from 10 minutes to 1 minute and built issue-discovery dashboards used by 40 analysts across a 200M+ member platform. Delivered a zero-incident cloud migration at ABN AMRO for a virtual assistant used by 22,000+ employees, and improved a model's accuracy from 50% to 97% through iterative training on an Upwork engagement.

Does Eugene have open-source work?

Yes — including an MCP server (wobbobobo-mcp-server), a Telegram bot that turns streaming song links into universal links (link-swapp), and a calendar-alert tool built on GitHub Actions and Twilio, all public on his GitHub.

Has Eugene led or mentored teams?

Yes — he led a cross-functional team including engineers, a data scientist, and a UX designer at ABN AMRO, and has mentored and onboarded new hires at both ABN AMRO and Netflix.

What's Eugene's educational background?

A Master's degree in Computer Science from Vinnytsia National Technical University (2011–2016), evaluated as equivalent to a Dutch Master's degree.

Does Eugene hold any professional certifications?

Yes — an Incident Responder certification from PagerDuty, dated October 2023.

What languages does Eugene speak?

Ukrainian, English, and Italian, plus basic Dutch.

How can I get in touch?

By email, LinkedIn, GitHub, or by booking a call directly — see the contact section below.

$ contact --init

Let's talk

Available for freelance and contract work. Get in touch to talk about your next AI or agentic system — or book a call directly.