Solve complex mongoDB aggregation pipeline task

I need someone who helps me to create a mongoDB aggregation pipline. I tried on myself, but mongoDB is new to me and writing brackets drives me crazy. So please - mongo gurus to the rescue! Find my attached data example with three records. The real database has now about 60.000 image records, but will soon have many hundred-thousands more. The collection is "images" and each image has many embedded "annotations". Annotations were made by internal staff members as well as external freelancers. The annotations of internal staff people are my ground truth for the comparison of the work of freelancers. I am collecting 2000 -5000 annotations from internal users, to create a "ground truth sample". Having this, i want to compare the results of external freelancers with the results from the internals to have a guess about the visual analysis ability of each external freelancer. In the first record, the image was annotated by user1@[login to view URL] and two external users user1@[login to view URL] and user2.extern.com. So, basically i need to check, if the image and its requested image feature was annotated the same way from the internal (master) and the external freelancer or differently. This needs to be expressed in some kind of a correlation value, or really simple, a percentage hitrate. I think the best would be to just receive an appropriate counter. I wanted to solve this with an aggregation pipeline, but i am stuck in all these matches, filters, projections and groupings. I am working on a Rails app using the mongoid gem, but i am fine with the mongo shell script as a result. I want to create a function in my app and send the user and the feature as parameters. Please check the attached files and when you are answering to my job post, please return the 4 magick words in the first line of the file [login to view URL] Looking forward your bids!