2.5x faster inference speeds with Trillium TPUs and Google Kubernetes Engine (GKE) for image generation
40% reduction in costs by optimizing AI/ML workloads with GKE
Less than 10-second turnaround time for user query response
20-30x faster model deployment time
HubX partnered with Google Cloud to deliver faster, more reliable responses to millions of users while keeping costs under control, ultimately achieving 2.5x faster performance at 40% lower cost.
HubX partnered with Google Cloud to deliver faster, more reliable responses to millions of users while keeping costs under control, ultimately achieving 2.5x faster performance at 40% lower cost.
If you've ever wished you could identify that mysterious plant in your garden or wanted to fix up a blurry photograph, HubX has a solution for you. As the makers behind some of the most popular apps worldwide — including DaVinci, Momo, PlantApp, PhotoApp, and more — HubX follows trending topics to create solutions for everyday challenges.
If you make a user wait 30 seconds from prompt to output, they won't wait. To reduce churn, everything needs to happen in under 10 seconds. That's the bar we're measured against — and it directly impacts the bottom line.
Cem Ortabaş
Co-founder, HubX
"Sudden spikes in search terms or online trends can instantly validate an app idea or render another obsolete, forcing developers to prioritize features, pivot strategies, and scale quickly to stay relevant," says Kaan Ortabaş, co-founder of HubX. "Generative AI is a top trend these days, so that influences the apps we develop."
However, in the mobile app market, performance is non-negotiable for success. "If you make a user wait 30 seconds from prompt to output, they won't wait," says Cem Ortabaş, co-founder of HubX. "To reduce churn, everything needs to happen in under 10 seconds. That's the bar we're measured against — and it directly impacts the bottom line."
That's why HubX uses Google Kubernetes Engine (GKE) to power its AI-driven applications, enable rapid development cycles, and tremendously improve the user experience.
Scaling up to meet user demands
Building highly scalable apps requires highly scalable infrastructure that balances speed, agility, and cost efficiency. With its previous setup, HubX experienced slow processing times and latency issues that caused high user churn, as well as difficulties with infrastructure management. This inefficient system got in the way of quick iteration and deployment, which is essential for staying ahead of customer expectations.
The company needed a scalable platform to quickly meet the demand of its AI/ML workloads. The team also wanted to improve developer productivity by empowering them with a faster, more intuitive development environment. Finally, the team aimed to optimize costs while maintaining high performance.
HubX's evaluation process led to Google Cloud and the unique capabilities of GKE. The platform's scalability, resource optimization features, and ease of use allowed HubX to efficiently deploy and manage its AI/ML workloads. Since some of HubX's developers had previous experience with Google Cloud, the switch was quick and easy, with workloads migrating directly from the previous environment into GKE. The team also benefited from the Google for Startups Cloud Program, which provided them with cloud credits, technical training and hands-on support, and access to a collaborative community of other startups.
Strong solutions for highly scalable AI applications
HubX now relies on GKE and AI Hypercomputer, using TPUs for fine-tuning and inference, A100 GPUs for larger, more complex models, and L4 GPUs for workloads that require fewer resources to optimize costs. This seamless TPU and GPU integration within GKE allows HubX to switch flexibly between different hardware accelerators as needed.
The company also uses Cloud Run and Cloud Run functions for rapid, serverless deployment of AI/ML applications, then migrates them to GKE for better management and control as they scale. "With GKE, we don't have to worry about the underlying infrastructure," Cem says. "This frees up our developers to build and deploy applications at scale."
With GKE, we don't have to worry about the underlying infrastructure. This frees up our developers to build and deploy applications at scale.
Cem Ortabaş
Co-founder, HubX
Meanwhile, Hyperdisk ML for GKE helps improve storage performance for HubX's AI/ML workloads, particularly when downloading large model images. "Handling large-scale machine learning models, especially under heavy customer loads, requires robust and high-performance storage solutions," Cem explains. "With Hyperdisk ML, we've unlocked impressive 20-30x faster boot-up times, ensuring that even the most complex ML workloads are ready to serve in record time."
Balancing speed, scalability, and savings
At HubX, we continuously push the boundaries of performance and efficiency. Google Kubernetes Engine and Google Cloud have been game-changers in this journey.
Cem Ortabaş
Co-founder, HubX
Since adopting GKE as the foundation for its applications, HubX has significantly reduced latency, with Trillium TPUs delivering 2.5x faster inference speeds. Users now get a response to queries in less than 10 seconds, improving user experience, conversion rates, engagement, and retention. HubX has also lowered its operating costs by 40%, allowing the team to allocate more resources to performance marketing to unlock future growth.
Using a managed platform like GKE also enables developers to iterate rapidly and keep pace with evolving trends. "We have significantly reduced cold start times by one full minute and improved inference speeds by 10% — bringing immediate and tangible improvements to our users," Kaan says.
Overall, GKE has given HubX the capacity to scale quickly and efficiently to handle fluctuating demands of its AI/ML workloads. This is especially valuable during peak usage times that are driven by marketing campaigns or trending applications. "At HubX, we continuously push the boundaries of performance and efficiency," Cem says. "Google Kubernetes Engine and Google Cloud have been game-changers in this journey."
HubX builds and publishes some of the world's most popular AI-powered mobile apps, including Nova, DaVinci, Tattoo.ai, and PlantApp.
Industry: Technology
Location: Turkey
Products: Google Cloud, AI Hypercomputer, Cloud Run, GPUs, Hyperdisk ML, Google Kubernetes Engine, TPUs
