Can I Run a Docker Container for Batch Processing from AWS Glue

Getting started with AWS Glue at my new workplace (I previously used AirFlow). My colleagues are running scheduled jobs via Glue, but they usually put a python script in the Glue editor, then adjust the schedule and some arguments. They cannot advise me how to get a finer control of the dependencies and binaries to control what kind of environment the script runs in. Also, it is not clear how does the script load any assets from disk (such as loading persisted machine learning models, which is my use case).

So, is there a way I can specify a Dockerfile which will build the Glue environment before running my script, or can I point to a GitLab repository containing the Dockerfile for this purpose?

Or else, is it possible to run a container itself fetched from Dockerhub? But then the Glue environment must have access to the Dockerhub registry, right?

To make the question concrete, consider the following script, that I can run via cronjob every Monday at 2 am.

#!/usr/bin/env bash
# encoding:utf-8

# Assume the host has docker installed 
# and docker login set up to access the container
# Run this script via Cron
function TIMESTAMP() { printf "%s" "$(date "+%Y-%m-%d %H-%M")";}
NAMESPACE=samy # Container registry in dockerhub, needs access to this
IMAGE="$NAMESPACE"/tf-clustering
CONTAINER=clustering-container
docker container stop $CONTAINER # Stop to enable deletion of the container
docker container rm $CONTAINER
docker image rm $IMAGE # To enable pulling the latest image from docker hub
set -e # From this point onwards, any failure must exit the script
docker container create --hostname $CONTAINER --name $CONTAINER $IMAGE
docker container start $CONTAINER
printf "%s %s Completed.\n\n\n\n" "$(TIMESTAMP)" "$CONTAINER"

Now, if instead of cron on a bare-metal (or virtual) ec2 instance, I need to run this in Glue, what are the necessary steps?