Company
ProperBird is a Munich-based prop-tech startup helping property portals accelerate growth via business intelligence. We specialise in deduplication and real-time crawling, are active in 25+ markets, and identify unique listings and agents.
Role overview
We're looking for someone to own how good our data is.
We collect property listings — and the agents and agencies behind them — from hundreds of websites across 25+ markets, and turn them into clean, comparable data our clients build on. The hard part is scale: nobody can check millions of records by hand.
So we're building automated checks that read through the data and judge whether each piece was read correctly. Some of it runs today. Most of it doesn't yet — that's the work of the coming months, and you'd be joining early enough to shape it rather than inherit it.
Until it's there, a great deal of the checking is done by hand, and that isn't a detour from the job. The judgements you and your team make are exactly what the automation gets taught from. Get them right and the system becomes trustworthy; get them wrong and it learns the wrong thing very efficiently.
The rest of your time — and over the first year this should become most of it — goes to the more interesting question: how do we make this better? Better at making sense of data that arrives in a different shape from every source. Better at telling when two records describe the same thing: the same flat advertised across five websites, or the same estate agent appearing under three different spellings. And better by bringing in outside data that sharpens both. This half of the job is open-ended, and the people who enjoy it tend to be the ones who are curious about why something was wrong rather than just fixing it.
Now and then a client notices something in the data and the question finds its way to you. First contact is handled elsewhere, so this is occasional rather than a standing commitment.
You will lead a team of two and report to one of our founders, taking the function over from them as they step back.
What you'll own
- How good the data is. Deciding what "correct" means for each kind of data, where the bar sits, and whether we are clearing it. Listings and agents are separate problems and get separate answers — being good at one tells you nothing about the other.
- Teaching the system. Establishing the right answers, correcting the automated checks where they get things wrong, and turning those corrections into the reference the system learns from — so the same mistake stops coming back.
- Making it better. The open half of the job: improving how we standardise data from very different sources, improving how we recognise the same property and the same agent across them, and finding external data that sharpens both.
- Making it scale. Sources keep being added, so each one needs to cost less than the last. Getting the automated checks to carry more of the load, and writing the work down well enough that someone new can pick it up, are both part of that.
- Two analysts. One already in the role, one joining at around the same time as you.
What you'll bring
- Experience: ≥5 years working hands-on with data — data engineering, analytics, data science, or a quality-focused role. You are technical enough to dig into the data yourself rather than asking someone else to.
- Real estate: you have worked with property data or inside the property industry — a portal, a listings or valuation product, an agency, a brokerage, a property manager, or anywhere else where the stock and the way it gets advertised were your daily material. You should already know why the same flat turns up on five sites with three different floor areas, and why that is not necessarily anybody's mistake.
- Curiosity about why data is wrong: the heart of this job is noticing the pattern behind a mistake and working out what would have prevented it.
- Comfort with unfinished ground: a good part of this is still manual and will be for a while. You need to be someone who does the work and builds the thing that replaces it, rather than waiting for the tooling to arrive.
- Judgement on messy data: real listings are inconsistent, incomplete and sometimes simply wrong. You can decide what the right answer is for an awkward case and explain your reasoning.
- Matching and deduplication: experience deciding whether two records describe the same thing — fuzzy matching, name matching, geographic data, or something similar. Matching people and companies is its own problem, and experience there counts as much as matching properties. You don't need to have built such a system, but you should understand how one goes wrong.
- Measuring rather than guessing: you're comfortable with accuracy measured properly — reference datasets, agreement between reviewers, precision and recall — and you know the difference between a number and an impression.
- Leading a small team: setting priorities for two people and keeping them unblocked, remotely.
- Writing clearly: definitions and processes that other people can follow without asking you.
- Professional skills: strong communication and teamwork; proficiency in English.
Nice to have
- A second European language, such as German, French or Spanish. It helps when checking a source in its own market, but it is not a requirement and we are happy to hire without it.
- Experience with geographic data, or with company and contact data at scale. Both turn up constantly in this job, and neither behaves the way people expect.
- Experience in more than one property market — the same field can mean different things in different countries, and that is most of what makes this hard.
- Experience reviewing or improving what an AI model produced.
Technical expertise
- SQL: strong — you can pull and check your own data without waiting on anyone.
- Python: comfortable writing scripts to pull, transform and compare data.
- Databases: MongoDB in production; ClickHouse and Postgres useful.
- Working with engineers: Git, reading and reviewing configuration files, and enough understanding of a data pipeline to tell where a problem started.
Benefits
- Fully remote — work from anywhere.
- Competitive salary paid in EUR.
- 25 vacation days per year.
Additional
- Start ASAP.
- 6-month probation.
- 12-month contract with annual extensions.
- Schedule Mon–Fri or Sun–Thu.
