EthoSafe needed a technology partner that was not only quick on its feet but had a deep understanding of our scaleable infrastructure needs. Agile Commerce provided us with the perfect combination of rapid development and thoughtful engineering. The team was creative, hard working and knowledgeable.
—Michelle Chambers, CEO
EthoSafe's founders are pioneers of social media. In 2008, they recognized that one of barriers to social media becoming a valuable business tool was the risk inherent with it. Free and open interactions of users and brands opens the possibility for someone to post inappropriate content that causes damage to a company's reputation and distress to their customers. EthoSafe was conceived to solve this problem.
The brief that Agile Commerce was given was to create a software-as-a-service solution that could easily be integrated into a social media site. It would provide accurate and rapid analysis of content to determine whether or not it conformed to the acceptable use policies of the site. It needed to incorporate artificial intelligence and human moderators. The average amount of time for a review was to be less than 30 seconds. Lastly, the solution needed to scale easily to accommodate spikes in intra-day traffic and overall growth of the popularity of social media and the size of EthoSafe's client base.
Our solution incorporated a simple XML-based API, similar to the APIs of Flickr and Facebook, that a site could call to start a content review. The core of the system was a series of orchestration servers running Windows Workflow Foundation to make enhancing the content review process simple. The heavy processing of content analysis was offloaded to an elastic cloud of servers running on Amazon Web Services. The system also incorporated artificial intelligence systems that were capable of learning. This allowed the system to increasingly leverage automated system for content review and depend less upon human moderators.
The final solution achieved or exceeded all of the parameters of the initial requirements. By using cloud computing for the computation-heavy processes and by using a distributed database structure, the system was able to scale gracefully and inexpensively in response to increasing demand. The orchestration servers were able to intelligently route incoming content for review. The average review time originally met the 30 second requirement, but quickly exceeded it by achieving average review times of under 10 seconds, with some pieces of content being reviewed in under a second.