Hi, I'm

Tomasz Sosnowski

Python Developer | AI & Automation

I design and deploy integration systems and automations β€” from architecture to production.

About me

πŸ‘¨β€πŸ’»

Solo IT in an e-commerce company β€” I design, build and deploy automation systems end-to-end, from marketplace integrations to AI agents. Outside work I build my own projects: exoplanet detection ML pipelines, NBA game predictors, AI email agents. I care more about understanding the problem than knowing the ready-made solution.

Technologies

PythonFastAPIOpenAILangChainQdrantDockerPostgreSQLRedisLangfusePresidioReactTypeScript

Work Experience

IT Specialist / Python Developer

January 2025 – present

E-commerce company Β· BiaΕ‚ystok, Poland

  • β–Έ Solo IT β€” I design, build and deploy systems end-to-end, from architecture to production
  • β–Έ Designed event-driven stock sync architecture: SQL Server Change Tracking β†’ RabbitMQ β†’ 15+ marketplaces and shops
  • β–Έ Built central product hub (FastAPI + PostgreSQL + Strapi) with offer export pipeline and BigQuery monitoring
  • β–Έ 36 integration repositories β€” order automation, stock sync, product images, reports
  • β–Έ All services containerized (Docker Compose), CI/CD via GitHub Actions, deployment via Portainer
View system architecture on GitHub

Projects

AI Support Mail Agent

RAG system for automated e-commerce customer email handling. Hybrid email classification (keyword matching + LLM fallback), local PII anonymization via Microsoft Presidio (no sensitive data sent to cloud), automatic PL/EN language detection, async processing via Redis queue, observability via Langfuse. 104 automated tests.

PythonFastAPIOpenAIQdrantRedisLangfusePresidioDockerGitHub Actions

Planet Hunter β€” Exoplanet Detection

Autonomous exoplanet detection system using NASA TESS photometric data. Multi-stage pipeline: light curve fetching from MAST API β†’ signal cleaning β†’ BLS periodogram β†’ signal classification. Custom decision-tree classifier distinguishes 7 classes (PLANET_CANDIDATE, ECLIPSING_BINARY, VARIABLE_STAR and more). Database of 476 known NASA planets for result validation. Processes thousands of stars 24/7.

PythonFastAPISQLiteDockerlightkurveBLSMAST APIML Pipeline

NBA Game Predictor

Machine learning system for predicting NBA game results (over/under total points). Automated daily data collection from NBA API β€” team stats, player injuries, schedule. XGBoost model with ~50-100 features, self-learning β€” trains on each completed game. Full backtesting and performance tracking.

PythonXGBoostPostgreSQLFastAPIDockerAPSchedulerSQLAlchemy

ElektroNova PriceWatch

Price monitoring system for electronics on Allegro, Amazon and Zalando with AI processing. Stealth scraping with anti-bot bypass, LLM result processing, price change notifications. Built as a demo for fictional company ElektroNova.

PythonFastAPIOpenAIDockerWeb ScrapingREST API

AI Automation Workflows

A set of business and AI automations on n8n platform. Check_Companies β€” weekly GPT scans stock market, evaluates AI/space companies via web search, saves results to PostgreSQL. IT Discovery Mail β€” delivers 3 IT/AI discoveries via email every morning. Watchlist Discord Digest β€” weekly investment report to Discord.

n8nOpenAIPostgreSQLDiscordResendDocker

Contact

I'm always happy to discuss collaboration or interesting projects.