About the job
As foundation models move from answering questions to acting over long horizons, memory becomes a central requirement. Systems must decide what is worth keeping and improve from experience. Large context windows and retrieval systems are not enough. You will work with researchers on projects across the company that require memory in different modalities to define general solutions that enable models to represent experience, decide what they keep and what they discard, and to efficiently store, retrieve and exploit memory.
Responsibilities
Design new memory architectures that work across problems, modalities and products
Define new ways to measure and evaluate memory performance
Help create and deliver production-ready, scalable and high-quality learning models and associated algorithms
Critically assess the effectiveness of such models and make recommendations for the ongoing roadmap
Qualifications
Minimum
PhD in Computer Science, Mathematics or related field, or 3+ years of relevant industry experience
Experience creating machine learning algorithms for vision and/or language problems and deploying them as production-level systems
Expert, hands-on knowledge of foundational models and vision and/or language, incl. local deployment and model post-training of open weights models
Expert, hands-on knowledge of memory, state, or long-horizon behaviour in foundation models (e.g., language agents, long-context methods, reinforcement learning under partial observability, world models, or sequence modelling)
Expert, hands-on knowledge of Python and associated ML tools/frameworks (numpy, scipy, sklearn, pytorch)
Experience with working collaboratively with other developers, following a flexible and agile approach
Preferred
No preferred qualifications listed.