Job Description
Kleinanzeigen reaches 36M+ monthly users and is Germany’s leading online classifieds market. We are a sustainability-focused platform where you won't just write code; you will power the services that bridge our high-traffic backend with seamless user experiences.
As a Data Engineer (Analytics & Products) at Kleinanzeigen, you start where the data need begins — not where the code does. You will work directly with product managers, analysts, and domain stakeholders to understand what decisions the data needs to support, how it will be consumed, and what already exists in the platform before writing a single line. From there, you design and build data assets that are homogeneous with the existing stack, cost-effective at scale, maintainable by the next engineer, and self-discoverable, so consumers can find, understand, and trust what you built without asking you.
You will own the full lifecycle: ingestion through Airflow DAGs, transformation across dbt stage, core, and report layers on Databricks, and the quality, documentation, and observability that make those assets production-grade. You stay current with Airflow and design DAG topology deliberately, knowing when to use dynamic task mapping, data-aware scheduling, the TaskFlow API, or a Cosmos dbt task group with Watcher mode for performance-critical pipelines. You understand Spark well enough to know why a dbt incremental model running in Databricks produces a full table scan instead of a partition filter push-down, how a poorly configured merge operation compounds into file fragmentation over time, and what to do about it, whether that means adjusting the incremental strategy, adding liquid clustering, running OPTIMIZE, or rethinking the model grain entirely. You write Python and shell scripts as naturally as SQL and follow engineering core principles: modularity, idempotency, testability, not because they are rules, but because they make your work last.
You operate in an AI-First engineering model: AI handles execution; you own intent, precision, and correctness. You use it to automate, to accelerate best practices, and to raise the quality bar across the team, not to ship faster with less judgment. You will collaborate closely with analysts, product managers, data platform engineers, and stakeholders across the company to ensure that the datasets we create and maintain in our data platform are reliable, trustworthy, and built to serve the decisions that matter.
What You Will Do
Data Modelling & dbt Development
- Implement dbt models across the medallion architecture applying the right materialisation strategy for each layer and use case — incremental, full refresh or snapshots — with consistent naming conventions, YAML documentation, metadata tagging, unit tests to validate critical business logic, and reusable macros for common transformation and data replication patterns
- Build and refactor models for different business areas of the company, including for example marketing performance, product metrics, C2C transactions, monetisation, vibrancy, and trust & safety
- Author reusable macros and apply consistent naming conventions, YAML documentation, and metadata tagging (billed, retention, gdpr)
- Design data models that are homogeneous with the existing stack, built at the right grain for the use case, cost-effective to run, and self-discoverable without needing the author to explain them
- Perform cost-aware modelling: clustering strategies, warehouse sizing, incremental scan reduction, and pre-aggregation layers
Data Ingestion & Pipeline Engineering
- Build and maintain Airflow DAGs using Python operators, designing DAG topology deliberately by choosing execution patterns, dependency structures, and sensor logic that match the operational requirements of each pipeline
- Design sensor logic for pipeline dependencies, including intraday vs daily completeness checks and DST-aware temporal handling
- Operate Cosmos dbt task groups within Airflow, including DAG splitting, warehouse selection, and Cosmos version upgrades
- Integrate new data sources end-to-end by building reliable, fault-tolerant connections across heterogeneous endpoint types including REST APIs, event streams, database connectors and file-based sources, with error handling, retry logic, and security patterns that make each integration production-safe from day one, unit testing every operator and transformation component where possible, and defining SLAs and SLOs that set clear expectations on data freshness, completeness, and availability for downstream consumers
- Work with data in the right format for each layer: Avro for event-driven ingestion schemas, Parquet for efficient columnar storage, and Delta for ACID-compliant lakehouse tables with time travel and schema evolution
Data Quality & Reliability
- Write dbt tests (not_null, unique, accepted_values) and semantic row-count checks to catch data quality issues before they reach consumers
- Author unit tests for critical business logic using dbt's unit test framework
- Own data quality across the full lifecycle: design and maintain quality check pipelines combining dbt tests, semantic validation and custom logic, monitor pipeline health proactively before issues reach downstream consumers, diagnose and resolve root causes across schema drift, silent failures, duplicate data and idempotency issues, and plan and execute backfills safely with pre-validation, before and after checks and documented rollback criteria
- Apply governance standards consistently across every pipeline and data asset you own, ensuring compliance requirements, data classification, retention policies and ownership accountability are traceable and maintained as the platform evolves
- Contribute to the reliability of the platform beyond your own work, flagging systemic issues, improving shared patterns, and leaving every codebase and dataset in better shape than you found it
Cost Optimisation
- Identify and implement cost reduction opportunities: warehouse downsizing, incremental model rewrites, DAG consolidation, unused table deprecation
- Ensure governance rules for compliance and cost ownership are applied consistently across pipelines and data assets — including tagging standards (business_unit, business_area, team) in Airflow and dbt
- Evaluate and act on compute cost signals from Databricks query cost tables
Stakeholder Collaboration & Knowledge Sharing
- Partner with analysts, product managers and domain teams to translate business requirements into accurate, reusable data models, and enable them to build and own their own data assets through patterns, templates, PR reviews and pair programming
- Provide consultancy and training to product teams on data ingestion and transformation patterns
- Support analysts in adopting dbt for report-layer development, including onboarding, PR review, and pair programming
- Enable teams to build their own data products: provide patterns, templates, and consultancy so analysts and product teams can develop and own report-layer assets independently
- Define and document reusable metrics, dimensions and business logic in the semantic layer, and capture institutional knowledge including domain assumptions, architectural decisions and data lineage context so it lives in the codebase and not in people's heads
- Participate in cross-team alignment on data model standards, shared definitions and platform evolution, representing the data engineering perspective in decisions that affect multiple domains
AI-First Engineering
- Review and verify all AI-generated code before submitting for peer review — you own correctness, not the AI
- Maintain machine-readable context files (CLAUDE.md) in repositories you own, capturing architectural constraints, naming conventions, and domain-specific rules for AI agents
- Use AI tooling as a force multiplier: author intent specifications, verify AI-generated outputs with full ownership of correctness, contribute to shared skills and context files, and use automation to raise quality standards across the team
- Build and iterate on Claude agents for scoped engineering tasks (e.g. report-layer model extension, documentation generation, comparison notebooks)
- Orchestrate multi-agent workflows for complex tasks (large-scale migrations, documentation backfills, cross-domain refactors) while verifying outcomes
- Contribute to the team's AI Skill Library: dbt patterns, ingestion templates, data quality check agents
What We're Looking For
Required
- Advanced SQL: window functions, CTEs, query optimisation, partitioning and clustering, with enough depth to understand execution plans and diagnose performance issues
- Hands-on production experience with dbt across the medallion architecture: models, tests, unit tests, macros, snapshots, YAML documentation and materialisation strategies
- Experience building and operating Airflow DAGs using Python operators, with the ability to design DAG topology deliberately for different pipeline requirements
- Proficiency in Python and shell scripting for pipeline development, ingestion tasks and operational tooling
- Solid understanding of lakehouse concepts and data formats: Delta Lake for ACID-compliant storage, Parquet for columnar efficiency, and Avro for event-driven schema definition
- Ability to work independently: owning tickets from design through delivery, including incident response and backfills
- Ability to gather requirements from non-technical stakeholders, understand consumption patterns, and translate business needs into data models that are accurate, maintainable and cost-effective
- Experience defining and maintaining schema definitions (Avro/ODIN or equivalent schema registry)
- Fluency with AI-assisted development workflows as part of daily practice: authoring intent specifications, verifying and owning AI-generated outputs, and contributing to shared context and tooling so the whole team benefits
- Experience with semantic layer concepts: defining reusable metrics, dimensions, and business logic that serve as a single source of truth across consumers (e.g. dbt Semantic Layer, Looker LookML, or equivalent)
- Spark comprehension at the level needed to understand how Databricks processes dbt models, read execution plans, and diagnose and fix performance issues such as full table scans, merge inefficiencies and file fragmentation
- Engineering fundamentals applied to data work: modularity, idempotency, testability and single responsibility as default practices, not afterthoughts
- Experience treating cost as a first-class constraint: identifying and acting on optimisation opportunities across compute, storage and pipeline design
Nice to Have
- Experience with Kafka or event-driven ingestion
- Familiarity with Salesforce, Google Ads, Meta, or other marketing/CRM APIs
- Experience with data modelling patterns and when to apply each: dimensional modelling, star schema, wide tables, one big table, slowly changing dimensions, and normalisation vs denormalisation tradeoffs
- Exposure to compliance requirements in data engineering: data classification, retention policies and privacy-aware modelling
- Experience with cost optimisation in cloud data platforms
- Experience orchestrating multi-agent AI workflows for complex engineering tasks
- Contribution to shared AI tooling, skills, or context files in previous roles
What Makes You a Great Fit
- Proactive by default: you don't wait to be asked. You flag issues before they become incidents, own follow-ups after delivery, and look for ways to improve things beyond the scope of your ticket.
- Generous with knowledge: you enjoy sharing what you know, whether that's through documentation, pair programming, code reviews, or supporting analysts and product teams in adopting the data platform.
- A team builder: you care about the people around you as much as the work. You contribute to a culture where the team gets better together, not just individually.
- An architect of intent, not just a code translator: you define the what and why with precision, use AI to handle the how, and verify that the output actually solves the problem. You understand that AI commoditizes syntax production — your value is in T-shaped thinking that connects business context, data modeling, and technical constraints.
- Quality-first mindset enabled by AI: you treat comprehensive testing, documentation, and data quality checks as non-negotiable — precisely because AI has made them cheap to produce. Low test coverage is no longer a time constraint; it's a choice.
Job Benefits
- An attractiveBase Salary
- Participation in ourShort-Term Incentiveplan (annual bonus)
- Work From Anywhere: Enjoy up to 20 days a year of working from anywhere! Maybe not from the moonwell why not! just make sure you have internet connection!
- A 24/7Employee Assistance Programfor you and your family, because we care
- Win together, lose together is one of our key behaviours. At Adevinta you will find a collaborative environment with an opportunity to explore your potential and grow
On top of these, we also provide a range of locally relevant benefits.
Adevinta is an equal opportunity employer and we value diversity. We do not discriminate on the basis of race, religion, colour, national origin, gender, sexual orientation, age, marital status or disability status.