If you're writing to MASTER and immediately reading from SLAVE, you will always run the risk of inconsistencies. You can mitigate the risk but you can never avoid it.
Doing everything possible to profile, tune and minimize the replication lag will help. Delaying the read query until the last possible moment will help. But it can't be avoided entirely if you're intent on never reading from MASTER.
If you time the replication lag under typical usage, you could configure something like django-multidb-router to read from MASTER for a specified period of time after write. Still is not 100% safe but you can configure it to be 99.9% safe for your setup.