Better Alternative to MongoDB db.collection.group() - mongodb

Question

I'm looking to group a collection of documents, then format them into something I can query dynamically, then query on the resulting collection. I've been able to group the collection and format it successfully using db.collection.group() but there are limitations when it comes to sharding that would prevent me from using it in production.
Here's that code:
var reduceFunction = function(current, result){
var provider = current.provider;
var collection = current.collection || provider;
result[provider] = result[provider] || {};
result[provider][collection] = result[provider][collection] || {};
result[provider][collection] = current.attributes;
}
var result = db.identity.group( {
key: { cookie: 1 },
cond: { organizationId: organization },
reduce: reduceFunction,
initial: {}
} );
Here's an example of the input documents:
{
"_id" : ObjectId("566c92712ae892aeb5278467"),
"provider" : "myProvider",
"collection": "myCollection",
"ip" : 111.111.111.111,
"cookie" : "12381902348asdf",
"updated" : ISODate("2015-12-12T21:32:33.824Z"),
"attributes" : {
company: "My Company"
}
}
Here's what my desired output would be (multiple input documents grouped by cookie)
{
"cookie" : "12381902348asdf",
"myProvider": {
"myCollection": {
"company": "myCompany"
}
},
"myOtherProvider": {
"myOtherCollection": {
"company": "myCompany"
}
}
}
so that way I could query using something like
collection.identity.find({"myProvider.myCollection.company": {$regex: /my/}})
This allows me to have conditions for multiple provider/collection/field groupings without having conflicts in the field names.