Machine Learning Research Scientist

Machine Learning Scientist (Probabilistic Inference)- AI Safety Research Non-Profit - Berlin (Hybrid)

We are seeking a Machine Learning (ML) Research Scientist specialising in probabilistic inference to join a highly technical research team working at the forefront of machine learning and probabilistic modelling. In this role, you will develop and evaluate advanced probabilistic inference methods, with a particular focus on amortised inference, translating theoretical insights into practical, scalable implementations.

Research & Algorithm Development:

• Develop novel amortised inference methods suitable for high-dimensional discrete and continuous probability distributions.
• Develop parameter-learning and structure-learning methods for large-scale probabilistic graphical models.

Evaluation & Experimentation:

• Design rigorous evaluation strategies for ML methods relying on probabilistic inference.
• Design and execute experiments to evaluate new inference and learning approaches, analysing results to drive future research directions.

Cross-Team Collaboration & Implementation:

• Collaborate closely with mathematicians and researchers on theoretical questions.
• Translate theoretical concepts and research proposals into high-quality, maintainable implementations using Python.
• Clearly communicate complex technical findings to multidisciplinary stakeholders.

Requirements:

• Advanced degree (PhD preferred) in Computer Science, Mathematics, Machine Learning, Statistics, or a related discipline.
• Minimum of 3 years of experience in deep learning or machine learning research, with strong expertise in probabilistic inference and a solid mathematical foundation.
• Track record of contributing to high-quality research in probabilistic inference, probabilistic modelling, or related fields.
• Experience in one or more areas: Bayesian inference, sampling-based approximate inference, amortised inference (variational inference, Generative Flow Networks), causal modelling, reinforcement learning, or optimal control.
• Strong Python programming skills and experience with ML frameworks like PyTorch or TensorFlow.