Steliau Technology

French distributor of electronic components managing 50,000+ product references across hundreds of manufacturer sources.

Visit website↗

When I joined Steliau, there was no technical profile in the company. I gathered requirements directly from operational teams, wrote the specifications, and built the PoC from scratch in 4 months. The goal: automate product data retrieval and structuring from the 400+ manufacturer websites Steliau works with. The obvious solution — one scraping script per manufacturer — doesn't scale. So I designed an adaptive AI scraper capable of automatically adjusting to any website interface without manual configuration. Technically: NLP models for named entity recognition (NER) and neural classification, an extraction pipeline, PostgreSQL database, FastAPI backend, and React interface. I also built a proprietary training dataset for the models. The PoC was demonstrated and validated up to the Executive Committee. This work later became SonarFetch — the fully productized version, built on personal time.