About IVORI Lab
IVORI Lab is an Oxford-based artificial intelligence research and product laboratory developing a Foundation Social World Model for decision-making under uncertainty.
IVORI Lab has closed its pre-seed round and is currently operating in stealth. It will formally announce its formation in October.
Founded out of the Oxford University Incubator, IVORI Lab brings together a multidisciplinary team spanning seven nationalities across three continents. Remarkably, nearly all founding and team members are under 25.
At the heart of IVORI Lab is a new approach to modeling human decision-making.
The laboratory's foundational model is being trained on centuries of humanity's highest-stakes decisions across politics, law, markets, the arts and science. It seeks to reconstruct historical counterfactuals — the hidden parallel worlds surrounding consequential decisions — and study how choices propagate through societies and across time.
IVORI Lab is initially deploying this technology in education, before expanding into applications in policymaking, capital allocation and scientific discovery.
The AI Alignment Research Team
IVORI Lab's AI Alignment Research Team is multidisciplinary. The team currently brings together researchers working across theology, philosophy, mathematics, computer science and machine learning.
The underlying research question is broader than simply building a better machine learning model:
How do we build artificial systems capable of forming judgment under uncertainty while remaining oriented toward human purposes and values?
We are approaching this question from first principles — combining normative, philosophical, biological, mathematical and computational perspectives to develop a coherent theory of alignment.
The Internship
This is not a conventional software engineering internship.
You will join IVORI Lab's AI Alignment Research Team as a Research Engineer, working on the mathematical and machine-learning foundations of a new AI alignment research program.
You will be joining a research team that is trying to build an alignment theory from first principles, and your role will be to help establish its mathematical and machine-learning foundations.
Your work will include:
- Reviewing technical literature on machine learning, world models, foundation models, representation learning, reinforcement learning, decision-making and AI alignment.
- Reviewing and analyzing emerging approaches to social world models and latent world modeling.
- Translating conceptual research questions into formal mathematical and computational frameworks.
- Exploring how decisions, actions, states, uncertainty and possible futures can be represented computationally.
- Investigating architectures and learning approaches relevant to modeling decision-making under uncertainty.
- Contributing to the mathematical and technical foundations of IVORI Lab's research program.
- Helping develop the technical background of IVORI Lab's concept papers.
- Identifying gaps and unresolved problems in existing machine-learning and world-model literature.
- Developing technical arguments, research notes and written sections for publication.
- Working directly with the research team to turn theoretical ideas into testable computational hypotheses.
Depending on your background and interests, the work may involve mathematical modeling, machine-learning theory, model architectures, representation learning, reinforcement learning, world models, simulation, evaluation or alignment methodology.
The precise scope will evolve with the research.
What You Will Be Working Towards
IVORI Lab is developing a Foundation Social World Model designed to model how decisions unfold within complex human environments.
This raises fundamental technical questions.
How should a model represent a decision?
How should it represent the state of a social environment?
How can it reason over counterfactual trajectories?
How can uncertainty over possible futures be represented?
How can a model distinguish between predicting what happened and understanding what could have happened?
And ultimately:
How can a machine-learning system develop representations that support judgment rather than merely prediction?
These are open research questions.
You will be working on the foundations of that problem.
Authorship
Where your intellectual contribution is substantive, you will be credited as a co-author of the relevant research paper(s).
This is intended to be genuine research experience rather than an internship where students are simply assigned implementation or background tasks.
Who We're Looking For
We are particularly interested in undergraduate students in mathematics, computer science, engineering, biology, medicine, physics or related quantitative disciplines, especially candidates from Oxford or Cambridge.
We are looking for students who are comfortable moving between mathematical abstraction, computational thinking and open-ended research problems.
What matters most is that you are technically strong, intellectually curious and interested in problems for which there is not yet a settled answer.
Why This Internship Is Different
You will be joining IVORI Lab at a particularly early stage: before the public launch, while the foundational research is still being written. You will not be handed an established architecture and asked to implement it.
You will be working alongside researchers asking what the architecture should be in the first place.
If you are interested in the future of AI — and particularly in the question of how increasingly capable machines can form judgment under uncertainty while remaining oriented toward human purposes and values — this is an opportunity to get unusually close to the research frontier.
To Apply
Please send a short note introducing yourself, your course and year of study, and one technical or abstract problem you have thought about recently that you found genuinely interesting, to [email protected].
We care more about how you think than how polished your CV is.
