Verschicke keinen 08/15-Lebenslauf — erstelle einen Lebenslauf und ein Anschreiben, die genau auf diese Rolle zugeschnitten sind.
Holidu in Munich seeks a Data Scientist to shape ranking and recommendation systems powering millions of guest searches. You will work end-to-end on models, from concept to production, with freedom to experiment and iterate.
The role emphasizes an AI-first approach, collaboration with product managers and engineers, and deploying with monitoring, A/B tests, and data-driven improvements. Proficient Python/SQL and large-scale data handling are required.
You'll join Search Intelligence, the team behind Holidu's ranking and recommendation systems. With over 8 million vacation rental properties on the platform, your models will determine which of them our 70+ million annual users discover first — directly shaping search conversion and business results. It's one of the highest-leverage problems in the company — and we're not just maintaining a finished system: we're pushing the boundaries of what search at Holidu can do. You'll work with a large and rich dataset, fast paths from concept to production, and the autonomy to act on your own judgment.
Holidu is an AI-first company, and it shows in how we work: your modelling stays grounded in machine learning and statistics, but you'll have large-scale access to the best AI tools and the freedom to weave them into everything you do — from discovery to implementation to maintenance. Everyone here experiments and shares what works, and we'll count on you to help push how far AI can take us.
This role is based in Munich with two office days/week.
Ranking at Holidu is already a deep system: a multi-stage pipeline with reinforcement-learning-based cold ranking, contextual re-ranking, and personalized recommendations, plus dedicated cold-start models for new properties with limited behavioral data. And it keeps expanding — we're starting to enrich our models with AI-generated semantic signals that describe what makes each home special, opening the door to smarter ranking and entirely new search experiences. The next generation of it is yours to figure out.
Some of the hard challenges we're tackling: balancing user relevance, conversion probability, and business value in a single ranking. Personalizing for users who are mostly anonymous or first-time visitors. Surfacing great new properties before any behavioral data exists — when a significant share of our inventory is new each quarter. And adapting our ranking to evolving partner and business models, where the right answer changes as the marketplace does.
Python Airflow dbt AWS (SageMaker, Redshift, Athena) MLflow
You'll shape the ranking and recommendation systems that millions of guests rely on end-to-end. You will: