Agent workflows
An n8n agent keeps a conference directory current: Claude Haiku re-checks tracked event pages weekly, and Claude Sonnet with web search finds new conferences monthly.
AI & Automation Projects
Agent workflows, LLM pipelines and internal tools that run on your real data, CRM, Jira and analytics. I have delivered LLM products since early 2023, and automated operations for years before that.
- 7M+
- records processed in lead-gen pipelines
- 1M+
- B2B leads processed for the DACH market
- 10+
- production cloud migrations
- 3,800+
- task specifications, now also for coding agents
Clients in AndorraWeb3 products · AustraliaInsurance reporting · AustriaMedical ERP · CanadaPRTS Media, Montreal agency (2018–2021), Upwork clients · ColombiaUpwork clients · CyprusWeb3 products · EstoniaProperty data pipelines, Telegram dating & events, Expense & receipt app · FinlandAI medical imaging, Expense & receipt app · GeorgiaWeb3 products · GermanyLead-gen data systems · IsraelFitness app, Upwork clients · South AfricaLanguage learning · UKNutrition coaching, Investment research, Upwork clients · United Arab EmiratesAdTech marketplace · USProject management SaaS, AI greeting cards, Upwork clients
What I do
LLM pipelines
A custom-trained GPT pipeline for company-data normalization, and an n8n and LLM workflow for multi-platform, multi-language publishing that replaced an external content vendor.
Data and scraping
Scraping, database sync, CRM enrichment, proxy/VPN reliability and Looker Studio reporting, with 7M+ records and 1M+ B2B leads processed.
Cost before hype
A lead-gen scraper started as a Claude agent, then moved to pure Python after I measured token cost against throughput.
Case studies
Lead-gen data systems
🇩🇪 Germany
First MVP: 2024
Coordinated scraping, database sync, CRM enrichment, proxy/VPN reliability, and Looker Studio reporting. Processed 7M+ records and 1M+ B2B leads for the DACH market.
300+ tasks • 11 contributors
Team mix: 5 backend, 2 frontend, 2 UI/UX, 1 DevOps, 1 QA.
AI medical imaging
🇫🇮 Finland
First MVP: 2023
Coordinated backend delivery for AI joint analysis of patients’ hand photos inside the client’s GitHub workflow: doctor/patient authentication, imaging history, ML orchestration, and an Azure database migration.
45+ tasks • 3 contributors • 4 departments
Team mix: 2 backend, 1 PM/BA.
Investment research
🇬🇧 UK
First MVP: 2021
3.5-year client relationship (2020–2023)
Defined a research dashboard and Stripe premium access for stock charts, earnings and signals. Kept client-supplied API data separate from data the team scraped and normalized.
Tools I built & operated
AI-powered SMM content pipeline
n8n plus LLM workflow for multi-platform, multi-language publishing with a group-command interface; replaced an external content vendor.
Agentic event-collection pipeline
n8n agent that keeps Approck’s conference directory current: Claude Haiku re-checks tracked event pages weekly, and Claude Sonnet with web search finds new conferences monthly. A Symfony API removes duplicates before results reach the directory.
Python lead-gen scraper
Started as a Claude agent, then rewrote it in pure Python after measuring token cost against throughput.
Marketplace operational tools
Python, n8n, BigQuery, and Data Studio workflows for app valuation, lead enrichment, contact deduplication, and pipeline reporting.