Modern science runs on tools as much as on ideas. Alongside our research, we build open-source infrastructure: frameworks, workflows and benchmarks that turn one-off analyses into reproducible, reusable capabilities.

We see software not as a by-product of research, but as part of its scientific core. A well-designed tool should make methods easier to inspect, compare, reproduce and extend. Each release is versioned, tested and documented, with the aim of creating capabilities that outlive the paper or project that first introduced them.

Our flagship software programme is CADAQUES, an open infrastructure for autonomous scientific discovery. CADAQUES is being developed as a common campaign layer connecting decision-making strategies to datasets, simulations, computational workflows and, ultimately, experiments.

We use this framework to study resource-aware benchmarking, Bayesian and agentic decision methods, multi-fidelity computation, asynchronous scientific workflows and the foundations of future autonomous laboratories.