In recent years, e-commerce has become widespread in society, and the logistics of product delivery is a crucial pillar for this market to maintain a high level of service and remain advantageous for consumers choosing to make purchases online. The present work aims to study the problem of last-mile vehicle routing for e-commerce deliveries and apply an Iterated Local Search (ILS) metaheuristic to optimize the routing of parcels in a Brazilian e-commerce company. With the objective of finding routes with the lowest cost for the deliveries, this study proposes an extension to the Vehicle Routing Problem with Occasional Drivers (VRPOD), considering a heterogeneous fleet and occasional drivers handling multiple deliveries. For the methodology application, data provided by an e-commerce company are used, and they are properly anonymized to prevent the identification of the company and its clients, respecting ethical principles. A total of 121 instances are used, ranging from the smallest with one vertex to the largest with 344. The results of the proposed model are presented in two scenarios: firstly, considering routing without the use of occasional drivers, and secondly, considering the availability of occasional drivers for some routes. Both scenarios are compared with the routes generated by the current router used in the company, and preliminary results indicate that without the use of occasional drivers, the proposed ILS obtains better solutions in 53.72 percent of the instances, and when occasional drivers are incorporated into the route, improvements occur in 76.03 percent of the instances. The utilization of occasional drivers also provides a 10.30 percent reduction in the average routing cost.