Full-stack marketer who builds the tracking, AI workflows and apps behind growth.
I run performance campaigns and build what they depend on: Tag Manager setups that send every conversion back to Google Analytics, Meta and Google Ads, AI automations in n8n, API integrations and custom apps. One person from the ad click to the CRM.
- Google Tag Manager
- GA4
- Meta CAPI
- Meta Ads
- n8n
- Claude
- REST APIs
- Next.js
01 / AI and automation
I automate the work that sits between your tools.
Most teams already have the software. What they lack is the glue: leads copied by hand, reports rebuilt every Monday, replies that wait until someone checks the inbox. I build that glue with n8n, direct API work and language models, then hand it over documented and running.
- 0.91Quote for 3 greenhouse unitsPetrakis AgroSales
- 0.84New lead: irrigation, LarissaMeta Lead AdsLead
- 0.62Where is my order #4821?Eleni VasileiouSupport
- 0.08Re: October product updateNewsletterLow
Custom AI workflows
n8n pipelines with AI agents that read, classify and route: lead scoring, inbox triage, content drafts, report summaries. Self-hosted or cloud, with logs you can actually read.
API integrations
REST, GraphQL and webhooks. I connect CRMs, ad platforms, messaging apps and internal databases, including the ones without a ready-made connector.
Works with your existing stack
No rip-and-replace. Automations plug into what you already use, so the team keeps its tools and loses the copy-paste.
- whisper-large-v3-turboReady
- llama-3.1-8b-instructIdle
- nomic-embed-textReady
Custom apps, local when it matters
Web dashboards, internal tools and native iOS or macOS apps. When data is sensitive, models run on the device through WhisperKit or Ollama.
02 / Tracking and performance marketing
I know which campaign brought the sale, because I set up the tracking.
I'm a full-stack marketer: I plan and run the campaigns, and I also build the measurement under them. Events defined on the page, routed through Google Tag Manager, and sent back to Google Analytics, Meta and Google Ads so their algorithms optimise for real conversions.
Google Tag Manager
Clean web and server-side containers: tags, triggers and variables named so the next person can follow them.
Event setup on the page
A written dataLayer spec, then the code to match: form submits, clicks, scroll depth, full GA4 ecommerce.
Sending conversions back
Meta Pixel plus Conversions API with event_id deduplication, Google Ads conversions and enhanced conversions.
Analytics you can trust
GA4, Matomo and Fathom, plus Looker Studio reports that show where the leads and revenue come from.
Running the campaigns
Meta Ads and Google Ads, landing page CRO and A/B tests, judged on the numbers the tracking returns.
page_view
page: /greenhouses
GA4Meta CAPIGoogle Ads
window.dataLayer.push({event: 'generate_lead',event_id: 'lead_8f3k2',// same id goes to the Conversions API, so Meta counts it onceform_name: 'quote_request',value: 120, currency: 'EUR'});
This site is set up the same way: server-side Tag Manager, a consent platform, plus Matomo and Fathom running alongside GA4.
03 / Featured build
KZ-SUBS
On-device subtitle generation for iPhone.
Import a video, transcribe it locally with Whisper, fix the captions on a timeline, style them and export. Built in Swift and SwiftUI with a strict MVVM split: heavy media and speech work runs off the main thread, and nothing ever leaves the phone.
- Speech model
- WhisperKit, large-v3 turbo, on the Neural Engine
- Languages
- Greek and English, word-level timings
- Editing
- Timeline, text and timing edits, karaoke styles
- Export
- SRT file or burned-in subtitled video
- Privacy
- No backend, no account. Works in Airplane Mode
04 / More work
Other builds
05 / Toolkit
Marketer and developer in one seat.
Knowing both sides means fewer handoffs. I can run the campaign, wire up the tracking, build the automation and read the numbers that tell us whether it worked.
Tracking and analytics
- Google Tag Manager
- Server-side GTM
- GA4
- Meta Pixel
- Meta Conversions API
- Google Ads conversions
- Matomo
- Fathom
- Looker Studio
Performance marketing
- Meta Ads
- Google Ads
- Conversion optimisation
- A/B testing
- CRM strategy
- SEO
AI and automation
- n8n
- AI agents
- Claude API
- OpenAI API
- Prompt design
- RAG and embeddings
- Ollama
- WhisperKit
APIs and back end
- REST
- GraphQL
- Webhooks
- OAuth
- Node.js
- Python
- Firebase
- PostgreSQL
Apps and front end
- React
- Next.js
- TypeScript
- Tailwind CSS
- Swift
- SwiftUI
06 / Journey
From game mods to AI agents.
2001
Frankfurt, the beginning
The game was Grand Theft Auto 3 and the hype was massive. I talked my parents into buying it, and after hours of playing one question kept coming back: how is something like this made? Without internet at home, the answer was hard to find.
2005
First computer
Television lost its appeal. I learned to search for everything: how is this built, how does that work. That led me to modding GTA San Andreas, building custom cars and models, and through modding I found programming. First C#, then Photoshop for the creative side.
2018
Web development
HTML, CSS and JavaScript. My first static sites and the fundamentals of design and modern development practice.
2019
React
Component-based architecture clicked. I started building interactive apps with real state management and more complex UX.
2020
Full stack
Node.js, databases and API design. End-to-end applications that pair clean interfaces with dependable back ends.
2021
Performance and SEO
Web performance, search and analytics. The focus moved to experiences that load fast and convert, not just look good.
2022
Marketing tech
CMS platforms, e-commerce and marketing automation. Combining technical skills with marketing strategy changed how I approach every build.
2023
Performance marketing
Became a Performance Marketing Specialist: running Meta and Google campaigns, and building the GTM setups, event tracking and conversion feeds that make them measurable.
2024
APIs everywhere
Integrations became the core of my work: CRMs talking to ad platforms, messaging apps talking to databases, all through their APIs.
2025
AI workflows
n8n pipelines with language models in the loop: triage, enrichment, drafting and reporting that run on their own. Plus local models with Ollama for work that has to stay private.
2026
KZ-SUBS
Shipped an on-device subtitle generator for iPhone in Swift, using WhisperKit for fully offline transcription.