ike GPS
ike GPS had a pretrained model and needed a solution to inference it while supporting auto-scale. There were also some latency issues, and they needed the resource code of the pipeline to support their model.
Arcanum provided a micro service that infers pole detection for IKE gps, by helping with productionising & optimising their current model solution.
We used our spellcaster to generate the code, post it to the Bitbucket to create the CI/CD pipeline and helped them provision the AWS infrastructure.
Guided by the metrics from our MLOps monitoring system, we found their model would benefit from a GPU, resulting in a 1000% performance increase.
We built them an auto-scale feature to ensure they could continue to process quickly as their demand grew, while keeping costs down during quiet periods.
We have successfully implemented a comprehensive solution to containerize the computer vision model for ike GPS, ensuring its efficient deployment as an API service.
AWS Solution Architecture
Leveraging the power of various AWS services, such as Route 53 for domain name management, VPC for secure networking, Internet Gateway for seamless internet connectivity, Load Balancer for distributing traffic, NAT gateway for enabling outbound internet access, ECS for container orchestration, ECR for secure container image storage, and EC2 for scalable computing resources, we have created a robust infrastructure.
This infrastructure allows us to seamlessly deploy the containerized computer vision model, providing a reliable and scalable API service to cater to the needs of IKE GPS. With this solution in place, we can confidently ensure the availability, performance, and security of the CV model as an API service, meeting the requirements of IKE GPS efficiently.
Overall it's been a pleasure working with Arcanum - We are in an interesting position of having extensive in-house AI experience, but not a lot of experience with running AI workloads on the Cloud. Arcanum have helped us to gain a better understanding of how we can best move our workloads to the cloud and have successfully implemented a solution for us on AWS that is performing very well.
Jeremy Gold, ike GPS, Principal AI / ML Engineer / Mad Inventor