Content understanding and embeddings are core to your work. You research and develop information-rich user and ad embeddings using unsupervised and generative approaches, directly improving the quality of predictions across both Playtime and adjoe Ads.

Raises the technical bar around them. You lead architectural debates, mentor junior scientists, and drive the team's research direction without needing a roadmap handed to you.

What’s the Team & Tech Stack?

The Data Science Research team builds foundational ML representations and advanced AI capabilities that enhance both Playtime and adjoe Ads. It leverages generative approaches to deliver systems such as high-dimensional user embeddings, general content understanding, and sequential recommendation. The stack focuses on deep learning and heavy data throughput: PyTorch and [Ray/Hugging Face] for training/fine-tuning foundation models, Trino and Athena for querying massive user-behavior data lakes, and dedicated GPU infrastructure.

Location & Benefits

We welcome applications from talent worldwide and provide relocation support to Hamburg, Germany for those ready to join our team.

What You'll Do

• Translate research into production-ready components for multi-stage recommendation pipelines, Generative Retrieval, Deep Ranking, and Re-ranking. • Develop and optimize embeddings and content understanding models. • Lead architectural discussions, mentor scientists, and shape the research direction. • Collaborate with cross-functional teams to deploy scalable AI capabilities in production environments. • Work with large-scale data from hundreds of millions of users to improve system performance and user experience.

Who You Are

• PhD in CS, Statistics, Math, Physics, or equivalent research depth with production experience; 4+ years in large-scale recommender systems. • Proficient in PyTorch; experience with model optimization, quantization, distillation, and deploying generative models in production. • Strong background in embeddings, content understanding, and unsupervised/generative approaches. • Proven ability to drive architecture, mentor others, and operate without a fixed roadmap.

Requirements

• PhD or equivalent research depth with production track record (or MSc with extensive industry research).
• 4+ years building large-scale recommender systems in high-throughput environments.
• Deep expertise in PyTorch and large-scale model optimization (quantization, distillation).
• Experience with Generative Retrieval, Deep Ranking, Re-ranking, and multi-stage pipelines.
• Strong knowledge of embeddings and content understanding.

Qualifications

• Publications in top-tier venues (SIGIR, KDD, NeurIPS, ICLR) or equivalent industry impact.
• Ability to translate research into production-ready code with low latency.
• Mentorship and leadership in research direction.