Senior Data Scientist - (Global Search, Consumer)

Deliveryhero; Berlin; Hybrid; EUR 90,000 - 130,000; Full time

Job summary

Delivery Hero is seeking a Senior Data Scientist to own the end-to-end ranking and learning-to-rank systems in our Global Search team. You will design, implement, and productionalize deep neural ranking models, ensuring scalable, low-latency inference across 60+ countries and 35 languages.

You will drive agentic ML research using LLM agents to automate feature engineering and experiment pipelines, mentor teammates, and collaborate with Backend/Data Engineers and Product Managers.

Qualifications

• Hands-on deep learning ranking architectures and LTR modeling.
• Experience with offline/online evaluation metrics (NDCG, MRR).
• Proficiency in Python ML stack and ML lifecycle tooling.
• Familiarity with agents/LLM tooling and autonomous experiment workflows.

Responsibilities

• Own the Ranking Stack end-to-end in production.
• Lead research into agentic development and autonomous experimentation.
• Collaborate with Backend/Data Engineers and PMs to deliver end-to-end ML improvements.
• Ensure model health: latency, drift, data quality in production.
• Mentor junior data scientists and promote best practices.

Skills

Deep Learning

Ranking & LTR

Python ML stack

MLOps

Experimentation

Cloud & Big Data

Education

Master's degree in CS/Math/Physics

Bachelor's with 6+ years experience

Tools

PyTorch

TensorFlow

scikit-learn

dbt

PySpark

Scala

Metaflow

Job description

We are on the lookout for a Senior Data Scientist - (Global Search, Consumer) to join the Search Ranking team within our Global Search tribe. If you thrive at the intersection of cutting-edge deep learning research and high-traffic production systems - and are excited about autonomous AI-driven experimentation - this role is for you.

Our search functionality spans more than 60 countries and 35 languages, facilitating over 80 million searches daily across four continents. The ranking systems you build will directly shape what millions of customers see every time they open the app.

Your mission:

• Own the Ranking Stack End-to-End: Design, build, and productionalize deep neural ranking models - including DCN-V2, MMoE, and Two-Tower architectures - operating at high throughput and low latency in production. You will own the full lifecycle: from offline experimentation and evaluation to monitoring, and iterative improvement.
• Drive Agentic ML Research: Embrace and champion the shift from manual experimentation to prompt-driven orchestration. You will leverage LLM coding agents (e.g., Claude Code, Gemini) to autonomously iterate on feature engineering and Learning to Rank (LTR) architectures. Inspired by the auto-research paradigm, you will design overnight experiment pipelines and review the outputs of hundreds of autonomous runs to identify signals quickly.
• Lead Feature Engineering and Model Architecture Innovation: Apply your deep expertise in ranking signals, feature stores, and embedding-based retrieval to push the quality of our rankers. Propose and validate new model architectures grounded in the latest research, translating academic advances into production-grade systems.
• Ensure Production Reliability: Maintain rigorous standards for model health in production. You will own monitoring for model drift, latency degradation, and feature pipeline integrity, and act quickly when signals deviate - keeping our ranking quality high for users across all markets.
• Collaborate Across Disciplines: Work as a technical partner with Backend Engineers, Data Engineers, and Product Managers to deliver end-to-end improvements. Translate complex ML trade-offs into clear narratives for non-technical stakeholders, and contribute to shaping the team's roadmap.
• Raise the Bar: Mentor junior and mid data scientists, drive best practices in experimentation rigor and code quality, and actively contribute to a culture of learning - especially around emerging agentic development workflows.

Master's degree (or Bachelor's with 6+ years of work experience) in Computer Science, Mathematics, Physics, or a related quantitative field. 4+ years of industry experience as a Data Scientist or Machine Learning Engineer applying ML in high-traffic production environments.

• Deep Learning Ranking & LTR: Hands-on, production-proven experience implementing modern deep learning ranking architectures - DCN-V2, MMoE, Two-Tower - with strong command of multi-task learning, cross-feature interactions, and embedding optimization at scale. Solid foundation in LTR methods (pointwise, pairwise, listwise), offline evaluation metrics (NDCG, MRR), and the challenges of bridging offline metrics to online A/B outcomes.
• Agentic & AI-Assisted Development: Proficiency with LLM coding agents and agentic CLI tools (e.g., Claude Code, Gemini) as first-class tools in your development workflow - not just for boilerplate, but for autonomous experiment generation, architecture prototyping, and rapid iteration. Familiarity with the auto-research paradigm and comfort reviewing large volumes of autonomously generated experiment results.
• Technical Depth & MLOps: Expert in Python and its ML ecosystem (PyTorch, TensorFlow, scikit-learn). Comfortable with SQL (dbt), cloud environments (GCP or AWS), and big-data processing at scale (PySpark, Scala). Strong command of the full ML lifecycle - feature stores, vector databases, model versioning, A/B experimentation frameworks, and deployment to high-throughput, low-latency serving infrastructure. Hands-on experience with MLOps practices: CI/CD for ML pipelines, Metaflow, experiment tracking, model registry, and production monitoring covering drift detection, latency tracking, and data quality alerting.
• Problem-Solving & Ownership: Comfortable navigating ambiguity and translating broad business objectives into well-scoped data science projects. You hold yourself accountable for outcomes, not just outputs.
• Collaborative Spirit: Demonstrated ability to work effectively in cross-functional, globally distributed teams. Strong communication skills to articulate model behavior, experiment results, and trade-offs to both technical and non-technical audiences.

Ensuring you and all our Heroes are looked after, happy, and healthy is always on the menu. Because if you're in good shape, then we're in good shape.

Make the most of our hybrid working model and join the team for face-to-face connection and collaboration in our beautiful Berlin campus 2 days