When using the myadroit.princeton.edu system (our Open OnDemand system) for a graphical Jupyter notebook session, the key to getting those custom conda environments to show up for your notebooks is to ensure that the ipykernel package (the IPython Kernel for Jupyter) is installed when you create your custom environment. For example,
conda create --name my_funky_env ipykernel scikit-learn matplotlib -c conda-forge
After building my new environment that way, the next time I launch a Jupyter notebook session from myadroit.princeton.edu I’ll see that new environment as an option when opening a new notebook from the “New” button on the upper right of the Jupyter window.