Independent AI engineering projects (document extraction, tool-calling agents, and RAG), built and evaluated end to end. Every project ships with its own evaluation harness, so the numbers below are measured, not estimated.

• Doc to CRM: reads invoices and delivery notes (PDFs or phone photos) into ledger entries, with a review UI for whatever it is unsure about. On 40 documents: 90% fully correct (100% on PDFs, 100% on clean photos, 60% on difficult ones). Human-review routing reaches 88.9% recall with zero false alarms and caught all 5 planted mismatches. Key finding is negative: the model's self-reported confidence was 1.00 on 36 of 40 documents, including ones it got wrong, so routing is driven by cross-checks instead.
• Inbox Agent: a tool-calling agent that triages a queue and stops before anything irreversible. Across 8 evaluation scenarios it picked the correct first tool every time and attempted 6 irreversible actions: 5 were surfaced for human confirmation, 1 was rejected by argument validation, none executed unattended.
• KB RAG Assistant: retrieval-augmented assistant over a private knowledge base, deployed and publicly reachable.
• Tech Stack: Python, FastAPI, Pydantic, pytest, Google Gemini, RAG, Docker, Hugging Face Spaces, GitHub Actions.

Built automation, integrations, and self-hosted Python infrastructure for a pan-European multi-brand e-commerce company, covering 11+ marketplaces and the full product lifecycle from sourcing to post-sale customer service.

Key achievements:

- Designed and maintained 350+ production workflows in n8n.io and Windmill: autoposting, order processing, stock/price sync, and error monitoring across 11+ European marketplaces (Bol.com, eBay, Mirakl-based: Hornbach, Empik, XXXLutz, BILD, Teract; Aukro.cz, Okazii.ro, OBI/VTEX, Metro DE, Praktiker.de, Pigu/PHH)

- Built and deployed 5+ production Python micro-services on a self-hosted Coolify platform (FastAPI/Flask + Playwright + Docker): replacing single-VM SPoF n8n workflows with per-service health-checks, cron scheduling, memory limits, and GitHub auto-deploy. Services include Metro DE marketplace publishing (HMAC-signed API, AI category resolver across 2179 leaf categories, 81-column PIM CSV uploads), VTEX UI automation with TOTP 2FA (auto-login via pyotp, persistent storage_state.json session), and stock-sync services with anti-bot IP probing and memory-safe Chromium launch profile.

- Re-architected legacy OBI/VTEX auto-posting pipeline from n8n to Windmill: a 15-step workflow with 4 cron schedules, AI category/dimensions/attributes/image-picker (Grok), Google-Sheets-backed state machine, and unified Bitrix24 error reporting.

- Integrated Odoo ERP (JSON-RPC) - Bitrix24 CRM - 10+ marketplace APIs into a unified data pipeline covering order fulfillment, invoicing, and delivery confirmation. Implemented custom x_auto_posting queue in Odoo with stage state machine and mail.message activity-log writeback for full audit trail.

- Developed eBay - Odoo two-way customer messaging: auto-fetching buyer messages, creating Odoo support tickets, and replying back to eBay, closing the full customer communication loop.

- Built a Pricing Intelligence System analyzing Bitrix24 order data: identifies negative-margin products, fetches live competitor prices via SerpAPI, applies a custom JS pricing algorithm to recalculate and update prices, and posts a structured HTML report back to Bitrix24.

- Implemented AI-powered categorization and content generation using OpenAI, Google Gemini, and xAI Grok (incl. Grok Vision for image scoring) with structured output, LOV-verbatim mapping for marketplace taxonomies, and fallback chains across providers.

- Architected AWS image pipeline (S3 + CloudFront + Thumbor) with content-hash deduplication, migrating from Cloudinary, which cut hosting costs by ~70% with marketplace-specific dynamic resizing.

- Established Bitrix24 error monitoring (dedup pattern, smart-deadline auto-assignment, RESPONSIBLE/ACCOMPLICES routing), which improved reliability by 40% and cut manual processing time by 60%.

- DevOps ownership: Docker multi-service builds, Coolify Apps + Scheduled Tasks, persistent storage, application secrets, healthchecks, Cloudflare-proxied SSL, Hetzner hosting, gradual rollouts, GitHub auto-deploy webhooks.

Tech Stack: n8n.io, Windmill, JavaScript, TypeScript, Python, Node.js, FastAPI, Flask, uvicorn, gunicorn, pydantic, pytest, Playwright (Chromium), Docker, Coolify, Hetzner, Cloudflare, GitHub Actions, Odoo (JSON-RPC), Bitrix24, VTEX (Catalog + Logistics + Seller Portal + 2FA), Metro DE, eBay, Mirakl, AWS (S3, CloudFront, IAM, Lambda), Thumbor, MongoDB, Google Sheets API, openpyxl, OpenAI, Google Gemini, xAI Grok (Vision), SerpAPI, HMAC-SHA256, OAuth, TOTP/2FA, REST APIs, Webhooks.

• Participated in a volunteer project, gaining hands-on experience with new technologies and delivering projects under tight deadlines.
• Developed user-friendly web applications using HTML, SCSS, and React.js, with backend development in Node.js.
• Managed project version control via Git and automated workflows with GitHub Actions, ensuring seamless collaboration.
• Ensured code quality with SonarQube and performed unit testing using Jasmine, Jest, and React Testing Library.
• Key technologies: SCSS, React.js, Node.js, MongoDB, Git, GitHub, SonarQube, Jasmine, Jest.

• Led end-to-end management of engineering projects, including cross-functional team collaboration, site visits, documentation review, and compliance assurance.
• Efficiently handled procurement, budget monitoring, and client/stakeholder communication, ensuring project success and satisfaction.
• Dedicated time to creating drawings, generating business offers, and maintaining communication with clients.