Junior data engineer - Italia
Pubblicato il 19-09-2026 - Cato in Varese
Your mission
Bring a tender from the source portal into Cato: scraping, parsing, merging, enrichment. You'll start by owning a handful of sources end to end - the scraper, the job behind it, and the data that comes out - and take on more as you go. Not tickets handed to you: sources you're responsible for.
What you'll actually do
- Build and maintain scrapers for national tender portals, where reading the source in its original language is part of the job.
- Keep them alive: portals change their HTML, move endpoints, break pagination, throttle you. You find out before the customer does.
- Turn messy sources into clean records: broken HTML, inconsistent XML, APIs that lie about their own schema.
- Write and maintain orchestrator flows: retries, backfills, alerting, and a clear answer to "Did today's run actually land?"
- Work on merge and dedup - the same tender arrives three times, in three shapes, and only one version can reach the customer.
- Ship AI enrichment steps: batch LLM extraction of requirements, embeddings, OCR on attachments.
- Guard data quality with tests and checks that fail loudly before a customer finds the gap.
Ideal profile
- Python that holds up: typed, tested, and readable six months later.
- You've scraped something real: HTTP, HTML and XML parsing, pagination, sessions, rate limits - and you know why a scraper that worked yesterday is broken this morning.
- SQL you're comfortable in: joins, aggregations, window functions. You'll read from the database every day; tuning and running it isn't your job.
- Builder by default: you see a manual process and your first instinct is to automate it.
- Comfortable with messy sources: broken HTML, inconsistent XML, PDFs that were scans of scans.
- You close your own loop: you check that what you shipped actually ran, before someone else has to ask.
Experience
- 1-2 years writing Python in production: scrapers, ETL scripts, automation - anything that had to run unattended and be fixed when it didn't.
- Exposure to an orchestrator (Prefect, Airflow, Dagster) is a plus, not a requirement: you'll learn ours properly.
- Exposure to LLM-based extraction is welcome; curiosity about it is mandatory.
What you won't find here
- No micromanagement: we trust you to own your part of the stack.
- No "standard" 9-to-5 mentality: we care about outcomes and we are looking for people who are willing to go the extra mile.
- No "we've always done it this way" excuses: we're here to disrupt, not to follow old patterns.
Our Tech Stack
- Data & Infra: Python, PostgreSQL, Prefect, AWS
- AI: batch LLM extraction, embeddings, OCR
Compensation
RAL €35,000 - €45,000 + equity, depending on profile.
Hiring Manager
Lorenzo Rossetto