diff --git a/.github/workflows/descriptions.yml b/.github/workflows/descriptions.yml
new file mode 100644
index 00000000..7d5ef504
--- /dev/null
+++ b/.github/workflows/descriptions.yml
@@ -0,0 +1,29 @@
+name: Update Docker Hub Description
+
+on:
+  push:
+    branches:
+      - master
+    paths:
+      - descriptions/*.md
+  workflow_dispatch:
+
+jobs:
+  dockerHubDescription:
+    runs-on: ubuntu-latest
+    strategy:
+      fail-fast: false
+      matrix:
+        include:
+          - image: cuda
+            short-description: NVIDIA CUDA libraries added to Rocker image.
+    steps:
+    - uses: actions/checkout@v2
+    - name: Docker Hub Description
+      uses: peter-evans/dockerhub-description@v2
+      with:
+        username: ${{ secrets.DOCKER_USER }}
+        password: ${{ secrets.DOCKER_PASSWORD }}
+        repository: rocker/${{ matrix.image }}
+        short-description: ${{ matrix.short-description }}
+        readme-filepath: ./descriptions/${{ matrix.image }}.md
diff --git a/descriptions/README.md b/descriptions/README.md
new file mode 100644
index 00000000..e32f8b8e
--- /dev/null
+++ b/descriptions/README.md
@@ -0,0 +1,3 @@
+# Pre-built images' descriptions
+
+Markdown files in this directory are descriptions of the pre-built images that are displayed on DockerHub.
diff --git a/descriptions/cuda.md b/descriptions/cuda.md
new file mode 100644
index 00000000..0ede6026
--- /dev/null
+++ b/descriptions/cuda.md
@@ -0,0 +1,71 @@
+# Rocker stack for Machine Learning in R
+
+`rocker/cuda`, `rocker/ml`, and `rocker/ml-verse` are Docker images for machine learning and GPU-based computation in R.
+
+These images are built in modular build system at <https://github.com/rocker-org/rocker-versioned2>.
+
+The dependency stack looks like so:
+
+```txt
+-| rocker/cuda
+  -| rocker/ml
+    -| rocker/ml-verse
+```
+
+All three are CUDA compatible and will optionally take R version tags (`rocker/ml:4.0.5`) with the option of additional trailing CUDA version tag (e.g. `rocker/ml:4.0.5-cuda10.1`).
+
+See <https://github.com/rocker-org/rocker-versioned2/wiki> for details of available tags and images.
+
+## Quick start
+
+**Note: GPU use requires [nvidia-docker](https://github.com/NVIDIA/nvidia-docker/)** runtime to run!
+
+Run a bash shell or R command line:
+
+```shell
+# CPU-only
+docker run --rm -ti rocker/ml R
+# Machines with nvidia-docker and GPU support
+docker run --gpus all --rm -ti rocker/ml R
+```
+
+Or run in RStudio instance:
+
+```shell
+docker run --gpus all -e PASSWORD=mu -p 8787:8787 rocker/ml
+```
+
+## Tags
+
+See [current `ml` tags](https://hub.docker.com/r/rocker/ml/tags?page=1&ordering=last_updated)
+See [current `ml-verse` tags](https://hub.docker.com/r/rocker/ml-verse/tags?page=1&ordering=last_updated)
+
+## Python versions and virtualenvs
+
+The ML images configure a default python virtualenv using the Ubuntu system python (3.8.5 for current Ubuntu 20.04 LTS), see [install_python.sh](https://github.com/rocker-org/rocker-versioned2/blob/master/scripts/install_python.sh).  This virtualenv is user-writable and the default detected by `reticulate` (using `WORKON_HOME` and `PYTHON_VENV_PATH` variables).
+
+Images also configure [pipenv](https://github.com/pypa/pipenv) with [pyenv](https://github.com/pyenv/pyenv) by default.  This makes it very easy to manage projects that require specific versions of Python as well as specific python modules.  For instance, a project using the popular `[greta](https://greta-stats.org/)` package for GPU-accelerated Bayesian inference needs Tensorflow 1.x, which requires Python <= 3.7, might do:
+
+```bash
+pipenv --python 3.7
+```
+
+In the bash terminal to set up a pipenv-managed virtualenv in the working directory using Python 3.7.  Then in R we can activate this virtualenv
+
+```r
+venv <- system("pipenv --venv", inter = TRUE)
+reticulate::use_virtualenv(venv, required = TRUE)
+```
+
+We can now install tensorflow version needed, e.g.
+
+```r
+install.packages("tensorflow")
+tensorflow::install_tensorflow(version="1.14.0-gpu", extra_packages="tensorflow-probability==0.7.0")
+```
+
+## Notes
+
+All images are based on the current Ubuntu LTS (ubuntu 20.04) and based on the official [NVIDIA CUDA docker build recipes](https://gitlab.com/nvidia/container-images/cuda/)
+
+**PLEASE NOTE**: older images, `rocker/ml-gpu`, `rocker/tensorflow` and `rocker/tensorflow-gpu`, built with cuda 9.0, are deprecated and no longer supported.