Staff ML Engineer - Search & Reco - m/w/d

AVIV; Berlin; Presencial; EUR 110.000 - 150.000; Jornada completa

Descripción de la vacante

AVIV Group is recruiting a Staff ML Engineer — Search & Reco (m/w/d) to own ranking and retrieval across AVIV's real estate platforms. You will design, train, and ship models for search relevance, ranking, and recommendations, from candidate retrieval through re-ranking and business-rule layers.

You will establish offline evaluations, design robust experiments, and drive the move toward LLM-augmented search.

Formación

• 8+ years of engineering experience and production ML systems.
• Experience with ranking, retrieval and recommendation systems.
• Ability to design offline/online evaluation and A/B experiments.
• Strong fundamentals: reliable data pipelines and reproducibility.
• Clear communication about uncertainty and decisions.

Responsabilidades

• Design, train, and ship models for search relevance, ranking, and recommendations.
• Own candidate retrieval, re-ranking, and business-rule layers.
• Set offline evaluation standards and run well-designed A/B tests.
• Ensure system architecture supports reproducibility and observability.
• Operate deployed systems, including monitoring and on-call rotation.
• Lead the move toward LLM-augmented search and semantic retrieval.
• Mentor teammates through design reviews and technical writing.

Conocimientos

Engineering experience

ML in production

Ranking systems

Evaluation methods

Communication

Herramientas

Vector search

RAG

Embeddings

Descripción del empleo

Staff ML Engineer — Search & Reco - m/w/d
Join one of the world's largest and most innovative digital real estate tech companies.

Backed by global investment leaders KKR and CPP Investments, AVIV Group is reshaping the way people across Europe buy, sell, rent, lease, and value real estate.

With a strong presence in Europe, we power some of the region's most recognized digital real estate platforms across various countries – including Immoweb, SeLoger, and Meilleurs Agents (France's leading real estate valuation platform), alongside immowelt and Homeday (one of Germany's fastest-growing PropTech companies). Our European footprint is complemented by our dynamic tech and engineering hubs in Romania and Croatia.

Our mission? We simplify life's most important transactions. By combining data, technology, and human expertise, we empower millions of people to find their perfect home!

Help us shape the future of real estate.

Role summary
You'll own ranking and retrieval problems across AVIV Group's real-estate platforms — from query understanding to re-ranking to production on-call. Every model decision ships with an explanation, and you're one of the people the team turns to when a metric win can't be explained.

What you'll do

• Design, train, and ship models for search relevance, ranking, and recommendations, from candidate retrieval through re-ranking to business-rule layers.




• Set the experimental standard: offline evaluation that predicts online behavior, well-designed A/B tests, and the judgment to kill a win that can't be explained.




• Own the architecture of the systems you touch — make build-vs-adopt calls, hold the line on reproducibility and observability.




• Operate what you ship: serving, monitoring, SLOs, and a share of the team's on-call rotation.




• Lead the move toward LLM-augmented search — query understanding, semantic retrieval, conversational surfaces — separating production-ready work from demos.




• Mentor through design reviews and technical writing that make good decisions repeatable.




• Work directly with product on what to optimize for, and push back with evidence when a target is measurable but wrong.

What you bring

• 8+ years of engineering experience, with at least 3 shipping ML systems to production at meaningful scale.




• Hands‑on experience designing ranking/retrieval/recommendation systems across the full funnel: candidate generation, feature engineering, model architecture, re-ranking.




• Experience diagnosing offline/online divergence and changing a team's evaluation practice.




• Strong engineering fundamentals: production-quality code, reliable data pipelines, reproducibility as a baseline.




• A track record of technical influence beyond your own output.




• Clear communication about uncertainty, including "we don't know yet."

Nice to have

• LLM-based search features in production (semantic retrieval, embeddings at scale, RAG, query understanding) with evaluation to back the claims.




• Learning-to-rank, vector search, or candidate-generation systems at high scale.




• Marketplace or classifieds domain background.




• Geospatial features in ranking.