Selected projects

Real automation systems, not demos. Click any card for full architecture, metrics, and tech stack.

Hiring signal:

Each project here shows how a business problem became a technical system — cost reduction, workflow automation, and production delivery across AI, B2B ops, and internal tools.

Discuss a project
RAG · AI Agent · E-commerce Production

Telegram Sales Agent — da-vinchi.pl

Production RAG assistant for a Polish wallpaper store. LangGraph ReAct + Hybrid RAG (SQL + pgvector). 355 SKUs, zero hallucinated prices, buyer-scenario tested.

Python LangGraph pgvector FastAPI Telegram
Key result: ~90% API cost ↓
View project →
Automation · LLM Pipeline Production

Invoice Automation Pipeline

Gmail → PDF → Gemini 2.5 Flash → Drive + Supabase. End-to-end invoice intake, validation, and structured storage. Weekly manual routine eliminated.

TypeScript Gemini 2.5 Flash GitHub Actions Supabase
Key result: Weekly routine → zero
View project →
Full-Stack · Multimodal AI Production

AI Image Composer

React 19 + Gemini multimodal web app for a creative studio. Replaced hours of manual prompting and mockup work. Identity replacement in 30 seconds.

React 19 TypeScript Gemini Canvas API Railway
Key result: Hours → 30 sec per artwork
View project →
Automation · B2B Ops Internal tool

B2B Lead Pipeline

Germany-focused outreach pipeline from Google Maps search to CRM-ready lead records. Automated company discovery, Impressum extraction, and lead enrichment.

Python Apify Gemini Supabase Pandas
Key result: 80-120 leads in 15-25 min
View project →
⚙

What these projects have in common

Business problem first. Every system here started as a manual bottleneck — a weekly routine, a multi-hour task, a process that depended on one person. The engineering goal was always the same: make it reliable, repeatable, and measurably faster.