A composable ecosystem for infectious disease modelling
Real-time infectious disease analysis usually forces a choice between two hard paths. You either chain separate models together, passing point estimates between them and losing the uncertainty along the way, or you write one large model tuned to a single problem that is hard to adapt or reuse.
EpiAware takes a third path. It is a set of small Julia packages that each do one thing — handling censored delays, convolving distributions, composing them — and combine into joint models that propagate uncertainty correctly. The package that ties them together, ComposableTuringIDModels.jl, lets you assemble a model from interchangeable infection and observation parts and turns it into a single Turing model you can simulate from and fit.
Where it is today
The ecosystem is at an early stage. CensoredDistributions.jl is the most developed and usable today; the joint-modelling and distribution packages are in active development, and some interfaces will change. Each package’s page says how settled it is.
Get involved
We are actively looking for collaborators — whether you contribute code, documentation, examples, or a modelling problem that composable tools could help with. See Get involved, or start a conversation on the epinowcast forum.