By using AWS, Visii runs its machine learning (ML) workloads on infrastructure that’s 3.5 times more cost-effective than its previous cloud environment. Its visual search technology provides an alternative to keyword search and broadens the search functionality of a client’s website. Visii uses Amazon Simple Queue Service (Amazon SQS) to manage thousands of users’ messages at a time to limit downtime, thereby increasing its agility.
Visii is pioneering visual search, a new way to browse sites online, with its machine learning (ML) technology. The service aims to help clients display the most relevant products to their customers and to improve the discoverability element of online browsing.
Visual search is an intuitive, online discovery and search tool that reduces search fatigue, where people struggle to articulate what they’re looking for in keywords. It can increase conversion rates by up to 30 percent by bringing more accurate products higher up in suggested images.
The London-based startup works with clients from a range of industries, from art to dating companies, as well as e-commerce retailers. Recently, Visii became an AWS Technology Partner in the Amazon Partner Network (APN).
Visii began as a part-time project researching and developing what became the core of their patent-pending product and technology platform. In late 2016, the startup was ready to go public with its visual search technology. Its initial infrastructure ran in a managed cloud, but Visii built its “messaging service” in-house because its provider couldn’t support such a complex environment.
This system controlled user image search requests, regulating the speed with which new searches came in to avoid overwhelming the servers housing its ML algorithms. With growth in mind, however, Visii wasn’t sure how much more traffic the system could handle. “Our provider experience at the time was antiquated and painful,” says George Whitelaw, chief technology officer at Visii.
"We could only reserve instances and buy computational time in advance, which was incredibly inflexible. It forced us to commit significant resources upfront for services we weren’t sure we’d still use in a few weeks’ time.” What Visii needed was a cloud environment that was agile, flexible, cost-effective, and friendly to startups.
A friend who understood Visii’s frustrations recommended to Whitelaw that he approach Amazon Web Services (AWS). After reviewing AWS solutions, the team quickly saw how Visii’s messaging service could work more efficiently in the AWS Cloud. Visii now uses Amazon Elastic Compute Cloud (Amazon EC2) for secure compute capacity and to run its public-facing APIs. It also takes advantage of Amazon Simple Storage Service (Amazon S3) for large-scale object storage.
Because Visii’s main priority is a flexible infrastructure, AWS recommended Amazon Simple Queue Service (Amazon SQS) to replace part of its messaging service. “This was mission-critical for us,” says Whitelaw. “By using Amazon SQS, we can maintain a sustainable pace of production as our ML algorithm analyzes images. It’s much simpler and cheaper than our previous setup.” Thousands of new images that would have jammed the production line are easily buffered by Amazon SQS, before images are filed using Amazon Relational Database Service (Amazon RDS).
Visii also relies on Elastic Load Balancing to effectively distribute incoming traffic across its Amazon EC2 instances and AWS Auto Scaling to monitor its applications, helping to ensure the system runs smoothly. This means Visii no longer needs to worry about scaling as it plans to take on new customers. Finally, Visii uses Amazon Route 53 to route user requests.
After moving to the AWS Cloud, Visii found its infrastructure to be 3.5 times more cost-effective to run, due in part to consolidating three Rackspace servers into Amazon SQS. Visii also uses the Amazon EC2 Spot Instance marketplace to keep compute costs low. “Since using Amazon EC2 Spot Instances, we’ve saved 45 percent on monthly compute costs, so we’re very happy,” says Whitelaw, “In fact, we slashed this figure in just seven weeks.”
Visii’s main aim, however, was to become a more agile business. “We’re now five times more agile as a company when we compare deployment processes required on AWS versus Rackspace,” says Whitelaw. “That means we now have the breathing space to foster creativity and focus on our core business.
The CTO also believes that the organization is now leaner. “DevOps used to be a complex task for one employee, but it is now shared among our developer team,” says Whitelaw. “We’ve removed the single point of failure."
Visii has been transformed by having access to AWS startup programs and the range of benefits they provide, including its own dedicated account manager. “Our AM has been paramount to our success,” says Whitelaw. “We’re constantly blown away by the amount of support we receive. AWS genuinely believes in us and champions us at every opportunity. Whenever friends approach me with an idea for a startup, I direct them to AWS.”
In fact, it was through its account manager that Visii applied and was accepted to the AWS Activate program for startups. It provided Visii with credits, training, and support to help the startup build its infrastructure in the AWS Cloud. Being selected to present as one of the seven most innovative startups at the AWS London Summit also helped it secure additional funding. Whitelaw says, “The AWS Activate program provided us with an environment in which we could experiment with minimal risk. We wouldn’t have had that opportunity with another cloud provider.”