diff --git a/doc/joss-2024/paper.bib b/doc/joss-2024/paper.bib
new file mode 100644
index 00000000..e69de29b
diff --git a/doc/joss-2024/paper.md b/doc/joss-2024/paper.md
index e69de29b..3bcc086f 100644
--- a/doc/joss-2024/paper.md
+++ b/doc/joss-2024/paper.md
@@ -0,0 +1,55 @@
+---
+title: 'genno: Efficient, transparent calculation on N-dimensional data'
+tags:
+- Python
+- energy
+- transportation
+authors:
+- given-names: Paul Natsuo
+  surname: Kishimoto
+  orcid: 0000-0002-8578-753X
+  affiliation: 1
+affiliations:
+- name: International Institute for Applied Systems Analysis
+  index: 1
+  ror: 02wfhk785
+date: 5 September 2024
+bibliography: paper.bib
+---
+
+# Summary
+
+Research in the fields of energy and transport systems, including integrated assessment modeling of energy and climate policies, often requires complicated manipulations of input and output data that are *multi-dimensional*, *labeled*, *sparse*, and represent many measurable quantities with multiple units of measurement.
+
+# Statement of need
+
+Code for handling such data can be fragile and opaque, which makes the validation, reproduction and extension of research difficult.
+In particular, adapting to new and revised input data, or refining model scenarios can involve extensive refactoring.
+
+`genno` is a Python package that builds on `dask`, `pandas`, and `xarray` to provide an API for transparent description and efficient execution of operations on multi-dimensional data.
+
+# Implementation and usage
+
+`genno` extends the `dask` directed, acyclic graph (DAG) data structure, which describe tasks and their inputs using Python types.
+While in `dask` these are used to distribute operations across multiple processes and nodes, in `genno` they are used to encode the data flow and manipulations in calculations expressed by the user.
+
+The user first *prepares* a ‘Computer’ containing description of many tasks, and then *executes* one or more tasks.
+`genno`, via `dask`, implements
+
+In preparing calculations, the user may use *keys* that allow concise but unambiguous reference to quantities to be computed.
+
+`genno` includes a large and growing library of fundamental *operators*, from which more complicated operations can be built up.
+
+# Example applications
+
+Two applications are described:
+
+1. Input data preparation for the MESSAGEix-Transport model.
+
+2. Integrated assessment modeling workflows.
+
+# Acknowledgements
+
+Many colleagues contributed to the initial design requirements for `genno`, including…
+
+# References