Methods and systems are described for encoder parameter setting optimization. A media item to be provided to users of a platform is identified. A request for content is received from a client device, and a media item associated with the content is identified. An indication of the media item is provided as input to a machine learning model. Outputs are obtained from the model identifying one or more sets of encoder parameter settings and, for each set, a confidence level that the settings satisfy a performance criterion based on the media item's media class. Based on the model outputs, at least one set of encoder parameter settings having a confidence level satisfying a confidence criterion is identified. The media item is encoded using the identified encoder parameter settings and provided for presentation via the client device.