Core
A common foundation for physical models, data and evidence. Geometry, parameters, provenance, assumptions and validity limits remain connected to each result.
calyr / CODE-BASED MODELS · SCIENTIFIC AI
calyr.aí
Surrogate models.
Better decisions.
calyr turns expensive physical models into fast, bounded decision tools. Surrogates compare alternatives; targeted simulations and experiments resolve uncertainty; assumptions, evidence and validity limits stay attached to every result.
Explore the platform01 / THE PLATFORM
One foundation.
Many scientific questions.
calyr Core, Engine and Model Capsules form one modelling architecture. Applications reuse that foundation while keeping their own physics, evidence and validation gates.
The aim is practical scientific inference: compare more alternatives, quantify what remains uncertain and choose the next simulation or experiment for information rather than habit.
Engine
Coordinates reference models, surrogate prediction, comparison and optimisation. Surrogates accelerate expensive calculations; the decision layer selects the next informative simulation or experiment.
Model Capsules
Reusable domain-specific model packages with explicit inputs, outputs, assumptions and evidence requirements. Each capsule is valid only for its declared use and tested domain.
Development status: a research and software platform under construction. Predictive performance and validation must be established separately for each model and use case.
02 / THE METHOD
Predict. Verify.
Decide from evidence.
A prediction becomes decision-relevant only when its assumptions, uncertainty, applicability and independent evidence are visible.
Define the question
Specify the decision, available evidence and constraints. Define what result would actually change the next action.
Predict and compare
Use physical models and calibrated surrogates to explore a bounded design or parameter space. Report uncertainty, applicability and out-of-domain cases.
Test what matters
Select the simulation or experiment that can discriminate between plausible explanations or reduce the decision-critical uncertainty. Verify predictions against independent evidence.
03 / APPLICATIONS
The platform,
put to work.
Research domains and application tracks test the same modelling principles at different physical scales and levels of maturity.
01aorta / aflowtexPatient- and device-specific aortic flow, mechanics and thrombogenicity modelling. A validation domain for multiscale inference, uncertainty and intervention design.02bríxGeometry, structural simulation and surrogate models for rapid engineering design comparison under explicit constraints.03biophysical inferenceSPR, SAXS and particle-scale models connect measurements to competing mechanistic explanations and test which model remains consistent with the data.
Further exploration: trajectoríes
An exploratory calyr track for path-resolved transport and purification design. The planned implementation uses differentiable computation to compare trajectories and operating conditions; LNP / mRNA is the first use case, with experimental validation still pending.
calyr develops scientific models, calibrated surrogates and inference. LITHÍOS applies that foundation to physical products, prototypes and design decisions.