Think about the structure of your S3 bucket and how you would later identify each object, which a user uploaded.
I would create for each user an initial key, which a user is able to upload the files, e.g.
username1/object1
/object2
/objectx
username2/object1
username3/object1
usernamex/objectx
This will give you the possibility, if a user is deleted, that you can just delete all objects with that username too. If you are using a generated key to identify the user, than you also can use the keyID instead of username.
The most interesting question is on how you will secure this, so that no other user will be able to see objects from others. If you have a underlying API, than it's "easy"... give the API the access to the S3 bucket and secure the requests, that only those objects will be listed for which the username or keyID matches.
If you are using IAM users (or roles), than you have automatically generate a policy for each base key (username1 or keyID) for the specific actions.
If you set up something like that, please be really sure to harden your security and also try to enable logging of this bucket to be sure, that user1 can't access objects from user2.
