Case Study: Skillup achieves scalable, reproducible ML for its training marketplace with Valohai

Version Control for ML as a necessity from the get-go

Skillup, a French SaaS marketplace for professional trainings, needed to keep thousands of course listings accurate while operating with a small technical team. They built web scrapers and machine learning models to identify training pages and extract fields (title, dates, program, duration), and partnered with Valohai to provide the version control, experiment tracking and orchestration required to run those models in production.

Valohai’s platform (with an API and an Airflow operator) automated experiment tracking, model/version storage, metrics reporting and cloud provisioning, letting Skillup run multiple classification models (page detection, field extraction, automatic categorization) reliably. The result: a three‑person tech team manages what traditional scraping would have required ~10 engineers, metrics and outputs are stored and shared automatically, cloud instances scale with a few clicks, and Skillup can now expand to thousands of sites and models.

Skillup

Ari Rouvinen

Head of data