Senior Research Scientist | Model Scaling (AI)

Location: Cologne, Germany; hybrid work with office attendance twice a week

Company

DeepL is a global AI product and research company building secure language AI solutions for translation, writing improvement, and real-time voice translation.

What you will do

• Select and evaluate open foundation and open-weight models for next-generation translation systems.
• Lead model selection and architecture decisions for scaling to hundreds of billions of parameters, including Mixture-of-Experts and other efficient designs.
• Design multi-capability adaptation strategies using LoRA, PEFT, and related methods.
• Own the modelling lifecycle from prototyping and ablation studies through scaling experiments, evaluation, and production delivery.
• Collaborate with post-training, RL/RLHF, and instruction-following specialists to integrate alignment and capabilities into base models.
• Track open-model and scaling research and translate findings into modelling recommendations.

Requirements

• Hands-on experience adapting and scaling multi-billion-parameter large language models through fine-tuning, instruction-tuning, or post-training.
• Strong understanding of architecture trade-offs at scale, including dense versus Mixture-of-Experts models.
• Working knowledge of parameter-efficient and multi-capability adaptation, including LoRA and PEFT.
• Experience training models, running experiments, debugging pipelines, and taking research results into production.
• Strong coding and experimentation skills with Python and PyTorch, JAX, or TensorFlow.
• Clear communication and effective collaboration across research, product, and engineering teams.

Nice to have

• Experience with uncertainty quantification, calibration, and confidence estimation for large models.
• Experience in machine translation, multilingual NLP, or document- and layout-aware modelling.
• Familiarity with MoE-specific training and adaptation, expert routing, Mixture-of-LoRA-Experts, and large-scale data-mixture design.

Culture & Benefits

• Internationally distributed team representing more than 90 nationalities.
• Open communication, regular feedback, and a collaborative growth-oriented environment.
• Hybrid schedule, flexible working hours, and coordination with team locations and time zones.
• Virtual Shares for every employee.
• Regular in-person team events and monthly full-day Hack Fridays.
• 30 days of annual leave excluding public holidays, mental health resources, and location-tailored benefits.