Challenge

The Artificial Intelligence / Machine Learning lab of this German telecom major was overwhelmed with a high volume of Proof of Concept and demo requests for AI / ML technologies from prospects and customers in Europe and North America. Slow execution of PoCs and demos was elongating sales cycles by delaying customer decision-making. Hiring or upskilling teams internally would increase project timelines and costs. The company anticipated a growing demand for AI/ML PoCs across different domains.

To augment its capacity, the company mandated its Global Capability Center to replicate its lab in India. The new AI/ML lab would mirror the hardware, networking, operating system and container environments of the German lab, and be in a position to create and execute PoCs at scale.

Solution

The GCC reached out to several OEMs and partners for help with setting up its AI/ML lab including HPE and Unique Solutions. We demonstrated extensive expertise in AI/ML infrastructure across leading OEM technologies like HPE Server and Nvidia GPU for high performance AI computing, SuSE Linux and Rancher Kubernetes for scalable orchestration, and Rancher Longhorn persistent storage solutions for stateful AI workloads. In addition, we showcased a wide array of certifications in deployment and administration of technologies of various OEMs. We backed up our capabilities with several success stories that showed our credentials.

We don't just sell boxes – we co-create solutions with our customers. Our team collaborated closely with the India GCC, understanding their AI/ML goals and architecting a scalable, future-ready AI Lab. We then provided a detailed approach for setting up the AI/ML lab and integrating it with the customer’s existing IT landscape. We then did a demo of SuSE Rancher / Kubernetes with stateless and stateful applications. Once the scope of work was finalized, we submitted a detailed proposal along with the bill of material of the required hardware and software components.

Impressed with our responsiveness and solution architecting expertise, the company selected us as its partner to set up its AI/ML lab.

Deploying a multi-vendor AI/ML infrastructure requires precision. After we supplied the ordered hardware and software, our engineers installed the HPE Server, Nvidia GPU, SuSE Linux operating system and SuSE Rancher container platform. We set up the required nodes and configured the pods. We also downloaded and installed the Jupyter Pod. We ensured seamless integration of compute, storage, containers, and AI software, as well as optimized Kubernetes cluster configurations for AI workloads.

We then handed over the system to the customer’s team for testing.

The key challenge in the implementation was integration of multivendor components with one another and of the entire lab environment with the customer’s network in India and Germany. Whenever there was an issue, we first triaged the problem and coordinated with the respective OEM to troubleshoot and fix it. Our technical expertise with all the involved technologies helped us localize the problem to server, storage, GPU, Linux, Rancher, or network access rights. As a top-tier partner of HPE, Nvidia and SuSE, we had access to the resources of all OEMs and used that to expedite the second step.

In the end, we delivered a scalable and high computing power AI/ML lab that was seamlessly integrated with the customer’s existing hardware and networking landscape.

At several stages during the engagement, we went the extra mile to support the customer. For example, we demonstrated Rancher Longhorn for persistent storage, finetuned Kubernetes configurations to enhance workload efficiency, and provided post-deployment knowledge transfer and best practices to the customer's engineering team.

Outcome

The customer reciprocated our partnership spirit by issuing a written appreciation for our engineering team.

The AI/ML lab is operational and running with efficient node management and orchestration of a multivendor landscape spanning compute, storage, container and networking. Operating at scale, it has slashed the lead time required to set up PoCs by half - from 6 weeks to 3 weeks – and has thereby accelerated the sales cycle. By shifting workloads from public cloud to onprem GPU-powered servers, the lab has cut PoC costs by 25%. In addition, the lab is fully equipped to handle diverse and growing AI/ML workloads, thereby future-proofing the customer’s investments.

Seeing the remarkable success of the AI/ML Lab, the customer is now expanding its capabilities further - demonstrating how Unique Solutions drives AI/ML innovation at scale.