On the Impact of White-box Deployment Strategies for Edge AI on Latency and Model Performance