A Cloud and Machine Learning-based Electronic Monitoring (EM) system that enables effective monitoring of the compliance of commercial tuna fishing was developed for NOAA Fisheries. HD video and GPS location trace data are pre-processed, with quality control via a scalable computing cluster Data Pre-Processing System, as part of a hybrid cloud solution for this EM system. The quality controlled video data and its metadata are ingested into Amazon GovCloud S3 storage. GovCloud elastic servers and other PaaS and IaaS services are used to facilitate EM data analysis and review. Machine learning-based Activity Recognition (AR) is another key solution in the EM system. AR detects and classifies fish capturing activities, tags the video segments containing fishing events, and performs video summarization to generate a compressed version of the video footage and other helpful artifacts. With the AR module, the amount of video data to be manually reviewed is reduced to 20% of the original data; the data analysis time is reduced by 40%, correspondingly. This comprehensive EM storage and analytics solution minimizes the data center investment, is capable of scaling out to other large EM projects, and cuts operational costs to a fraction of the cost incurred with the traditional approach.