I sync data between my local DB and a Server.

I'm looking for the cleanest way to modelise all of this.

I have a com.something.db package That contains a Data Helper and couple of DAO classes that represents objects stored in the db (I didn't write that part)

com.something.db
--public DataHelper
--public Employee
         @DatabaseField e.g. "name" will be an actual column name in the DB
         -name
         @DatabaseField
         -salary
         etc... (all in all 50 fields)

I have a com.something.sync package That contains all the implementation detail on how to send data to the server. It boils down to a ConnectionManager that is fed by different classes that implements a 'Request' interface

com.something.sync
--public interface ConnectionManager
--package ConnectionManagerImpl
--public interface Request
--package LoginRequest
--package GetEmployeesRequest

My issue is, at some point in the sync process, I have to JSONise and de-JSONise my data (E.g. the Employee class).

But I really don't feel like having the same Employee class be responsible for both his JSONisation and his actual representation inside the local database. It really doesn't feel right, because I carefully decoupled the rest, I am only stuck on this JSON thing.

What should I do ?

Should I write 3 Employee classes ?

EmployeeDB
   @DatabaseField e.g. "name" will be an actual column name in the DB
   -name
   @DatabaseField
   -salary
   -etc... 50 fields


EmployeeInterface
   -getName
   -getSalary
   -etc... 50 fields


EmployeeJSON
   -JSON_KEY_NAME = "name"   The JSON key happens to be the same as the table name, but it isn't requirement
   -name
   -JSON_KEY_SALARY = "salary"
   -salary
   -etc... 50 fields

It feels like a lot of duplicates.

Is there a common pattern I can use there ?