Menu

Projects

What we are working on

Each of these is a question DAIM has committed to: what we are asking, which themes it draws on, and what is being done about it now.

  • 01

    Four time-series traces of different character converge into a single model, which produces a shared representation. A dashed path indicates transfer to domains not seen during training.

    DAIM Foundation Models

    Cross-Domain Foundation Models for Time-Series Discovery

    Can a foundation model learn dynamics that transfer across domains?

    Investigating general-purpose models for time-series data across science, health, engineering and society, starting from reproduction and systematic evaluation of the models that already exist.

    Current focus Cross-domain benchmarking, synthetic dynamics and transferable representations.

    Active Research Foundation ModelsTime Series

    Frontier AI

  • 02

    Literature, data, experiments and human decisions feed intermediate agent steps that lead to a single claim. A dashed return path indicates that the claim can be traced back to its sources.

    DAIM Research Agents

    Evidence-Grounded Agents for Reproducible Research

    Can AI agents contribute to research without losing the link between claims and evidence?

    Investigating agentic AI systems that support literature analysis, experimental design, coding, reproducibility, scientific critique and research synthesis — with conclusions that stay traceable to their sources.

    Current focus Research workflows, evidence provenance, reproducibility and human–AI collaboration.

    Active Research Agentic AIAI for Research

    Frontier AI

Proposing a project

A project usually starts much smaller than it looks here — a question someone brought to a seminar, or an idea that held up once other people pushed on it.

If you are inside DAIM, that is the route. If you are outside it, get in touch and we can talk about it.