By using Cyclops on the AWS Cloud, our customer now has the data to leverage the new wave of experience-based retail without spending millions of dollars on data science.
Jung Hong Kim CTO
  • About Dayta AI

    Hong Kong–based startup Dayta AI has developed Cyclops, a software-as-a-service machine-vision technology that helps retailers identify customer behavior patterns from security camera images and leverage customer experience management.

  • Benefits of AWS

    • Avoids upfront hardware costs
    • Offers machine-vision technology at a compelling price point
    • Gains up to 20% greater computing price-performance
    • Enables customers to leverage experience-based retail
    • Stays focused on ML modelling
  • AWS Services Used

Machine vision, sometimes referred to as computer vision, enables computers to identify patterns in image data captured by cameras. Hong Kong–based startup Dayta AI has developed Cyclops, a machine-vision software that retailers use to learn more about their customers’ retail experiences. Cyclops machine-vision technology identifies patterns such as the demographics of shoppers, their emotions, and their journeys through the stores. Retailers then use the data to optimize their operations such as store layout, staffing levels, and in-store or offline marketing.

With its limited resources, Dayta AI wanted an infrastructure for Cyclops that could be scaled without any upfront capital expenditure. It also sought an infrastructure that wouldn’t be costly to manage. Patrick Tu, CEO of Dayta AI, says, “A cloud infrastructure would help us keep our costs down because of the cloud’s pay-as-you-grow model. Plus, we could offload routine infrastructure management tasks to the cloud provider.”

Dayta AI assessed leading cloud providers including Amazon Web Services (AWS), Google Cloud Platform (GCP), and Microsoft Azure. “We saw that Cyclops performed better on AWS than other cloud services,” says Jung Hong Kim, CTO of Dayta AI. Cyclops software runs in containers managed by Kubernetes. Kim says, “Kubernetes on AWS was solid. There were no slowdowns, and we could scale the containers up and down really smoothly.” Dayta AI found that Amazon Elastic Compute Cloud (Amazon EC2) instances were also highly compatible with open-source frameworks crucial to operating Cyclops using Kubernetes. “It allowed us to deploy and scale our software without any barriers,” Kim says. 

When comparing cloud providers, Dayta AI looked at the high-performance graphics processing units (GPUs) for machine learning (ML) algorithms. Says Eugene Ho, chief data scientist of Dayta AI, “We discovered that Amazon EC2 P3 instances provided up to 20 percent better price-performance than comparable services from other cloud providers in terms of teraflops-per-hour-per-US dollar. With AWS, we gained the machine learning processor performance we needed within our budget and could offer customers our machine-vision technology at a compelling price point.”

Dayta AI joined the AWS Activate program and gained promotional credits for the development of Cyclops. The company also had access to AWS Solutions Architects, who advised Dayta AI on what databases to use. Kim comments, “Choosing the right databases for Cyclops was vital because of our need for low-latency performance. We had several options, and an AWS Solutions Architect helped us to choose between a SQL or No-SQL database, and whether to place it inside or outside our Kubernetes containers.”

Dayta AI decided on Amazon DynamoDB as the database service for Cyclops, operating inside an Amazon Virtual Private Cloud (Amazon VPC) network, which provides single-digit millisecond performance at any scale. Kim comments, “Amazon DynamoDB gives us the throughput we needed for Cyclops to collate the gigabytes of data pouring in from customers’ in-store cameras.” Anonymized and aggregated data from the cameras is stored in Amazon DynamoDB after being processed by multiple ML models running on the Amazon EC2 P3 instances. The ML models include a journey-mapping model and a heatmap-generation model, which help retailers understand customers’ movements, emotions, and demographics.

Dayta AI automates the building and development of Cyclops and its many ML models using AWS CodeBuild and AWS CodeDeploy, while AWS PrivateLink ensures the security of data shared with cloud-based applications. “By leveraging various AWS services, we can protect our intellectual property as well as our clients’ data without costly and time-consuming configurations. It means we stay focused on developing our ML models,” says Kim.

Cyclops running on AWS has already helped a leading global distiller of cognac measure the impact of its event-marketing activities on customers in China and Southeast Asia. The distiller runs quarterly exhibitions to promote the brand and history of cognac and engaged Dayta AI to analyze the experience of visitors at each event.

Key findings from Cyclops included identifying the discrepancy between the intended and actual customer experience in several sections of the exhibition. For example, a section with a 3-minute video of the history of cognac was visited for only 30 seconds on average. The company redesigned the section, adding immersive displays and chairs, resulting in a 200 percent increase in the time people spent watching the video and an improvement in visitors’ emotions metrics.

Kim says, “Our client now has the information to create data-driven customer experiences and is trying different combinations of staff allocation to optimize employee experience as well. By using Cyclops on the AWS Cloud, our customer now has the data to leverage the new wave of experience-based retail without spending millions of dollars on data science.”

To learn more, visit or contact us at