Tom Kraljevic,
VP of Engineering
VP of Engineering
"Instead of waiting 6 months for the customer to procure hardware for a POC, we can deliver them something quickly. They can just plug [the appliance] in and then be able to work with it. Not only does it speed up and simplify the environment, and reduce the need for technical people on the customer side, it just speeds everything up."
Watch the Video
Company Size:
400
Industry:
Artificial Intelligence
Challenge
H2O.ai is an AI and machine learning platform that provides open-source and enterprise solutions for building, deploying, and automating AI-driven insights and applications.
H2O.ai needed to deliver a complete AI appliance—hardware and software combined—to their customers operating in highly secure, on-premises environments. These customers, including government agencies, banks, and telecom providers, require strict security measures, often working in air-gapped environments where internet access is either restricted or entirely unavailable. This posed a unique challenge: how to provide a powerful AI solution that could function seamlessly and ensure compliance with rigorous security policies without relying on external connectivity.
Deploying AI in these settings is further complicated by the need for specialized infrastructure. The AI appliance had to include the necessary compute resources, such as GPUs, while integrating Kubernetes orchestration in a way that did not require ongoing management from the customer. Additionally, given the security-sensitive nature of these deployments, software updates and maintenance had to be handled in a controlled manner without introducing vulnerabilities.
Another major consideration was deployment efficiency. Traditional IT procurement cycles can delay AI implementation by months, as customers must first acquire and configure compatible hardware before installing software. H2O.ai aimed to eliminate these bottlenecks by delivering a fully integrated AI appliance that could be quickly deployed, ensuring customers could start leveraging AI capabilities with minimal setup time and technical overhead.
When H2O.ai discovered Replicated’s Embedded Cluster, they found a perfect match for what they wanted to build.
H2O.ai needed to deliver a complete AI appliance—hardware and software combined—to their customers operating in highly secure, on-premises environments. These customers, including government agencies, banks, and telecom providers, require strict security measures, often working in air-gapped environments where internet access is either restricted or entirely unavailable. This posed a unique challenge: how to provide a powerful AI solution that could function seamlessly and ensure compliance with rigorous security policies without relying on external connectivity.
Deploying AI in these settings is further complicated by the need for specialized infrastructure. The AI appliance had to include the necessary compute resources, such as GPUs, while integrating Kubernetes orchestration in a way that did not require ongoing management from the customer. Additionally, given the security-sensitive nature of these deployments, software updates and maintenance had to be handled in a controlled manner without introducing vulnerabilities.
Another major consideration was deployment efficiency. Traditional IT procurement cycles can delay AI implementation by months, as customers must first acquire and configure compatible hardware before installing software. H2O.ai aimed to eliminate these bottlenecks by delivering a fully integrated AI appliance that could be quickly deployed, ensuring customers could start leveraging AI capabilities with minimal setup time and technical overhead.
When H2O.ai discovered Replicated’s Embedded Cluster, they found a perfect match for what they wanted to build.
“When the hardware appliance idea came along, it felt like a perfect match for the Embedded Cluster itself. From the development and experience standpoint, everything about Embedded Cluster is very streamlined. It’s small and lightweight.”
Tom Kraljevic,
VP of Engineering
VP of Engineering
Solution
H2O.ai partnered with Replicated to leverage Embedded Cluster, allowing them to package Kubernetes, management tools, and AI software into a single deployment. H2O.ai’s AI appliance arrives to their end customers pre-configured with a 2U rackable server, NVIDIA GPUs, and H2O.ai’s AI platform, ensuring a seamless out-of-the-box experience.
By delivering a fully integrated AI appliance, H2O.ai significantly shortened its sales and onboarding cycle. Instead of waiting for customers to procure hardware before even beginning a proof of concept, H2O.ai could now provide a ready-to-use system immediately. As Tom Kraljevic, VP and Fellow at H2O.ai, put it, "instead of waiting six months for the customer to procure hardware just to do a proof of concept, we can deliver something quickly. They just plug it in and it’s ready to go. It speeds everything up and reduces the need for technical personnel on the customer’s side."
Replicated’s Embedded Cluster enables vendors to ship their application and Kubernetes together as a single binary, simplifying how applications are packaged for self-hosted deployments in secure enterprise environments. This streamlined approach eliminates the complexities of infrastructure management - for both H2O and their end customers. "Not only does the customer not need to think about Kubernetes, but I don’t have to think about it either," Kraljevic noted. "It just works in the background, and the entire system gets updated seamlessly."
The result is a secure, efficient, and easy-to-deploy AI solution. Customers now receive a ready-to-use appliance that accelerates their AI adoption while ensuring reliability in even the most demanding environments. At the same time, H2O.ai has drastically improved deployment speed, security, and user experience.
Download Case StudyBy delivering a fully integrated AI appliance, H2O.ai significantly shortened its sales and onboarding cycle. Instead of waiting for customers to procure hardware before even beginning a proof of concept, H2O.ai could now provide a ready-to-use system immediately. As Tom Kraljevic, VP and Fellow at H2O.ai, put it, "instead of waiting six months for the customer to procure hardware just to do a proof of concept, we can deliver something quickly. They just plug it in and it’s ready to go. It speeds everything up and reduces the need for technical personnel on the customer’s side."
Replicated’s Embedded Cluster enables vendors to ship their application and Kubernetes together as a single binary, simplifying how applications are packaged for self-hosted deployments in secure enterprise environments. This streamlined approach eliminates the complexities of infrastructure management - for both H2O and their end customers. "Not only does the customer not need to think about Kubernetes, but I don’t have to think about it either," Kraljevic noted. "It just works in the background, and the entire system gets updated seamlessly."
The result is a secure, efficient, and easy-to-deploy AI solution. Customers now receive a ready-to-use appliance that accelerates their AI adoption while ensuring reliability in even the most demanding environments. At the same time, H2O.ai has drastically improved deployment speed, security, and user experience.
"As a company, our customer base has always been air-gapped… the appliance thing is taking it to the next level. By using Embedded Cluster and providing the actual hardware envelope, we don’t need to wonder what the customer environment is going to be. It’s going to be correct by definition."
Tom Kraljevic,
VP of Engineering
VP of Engineering
%201.png)