Three hours — this is exactly the time the artificial intelligence needs to process a database of 1,844 contacts. The same amount will take a person over 30 working days. Here’s how a GigaCloud’s GPU-enhanced cloud helps UniTalk develop AI-based solutions and simplify the work of its customers.

UniTalk is an international telecommunication IT company. It develops the platform for business communications of the same name, which includes: IP telephony and virtual PBX, call tracking, recording of phone conversations, website callback buttons and widgets, various types of phone calls and routing, etc. Clients include businesses of all size and the public sector.

Looking for a cloud operator

Since 2014, the UniTalk company has been providing uninterrupted communication and high-quality telephony to customers in more than 12 countries around the world. Since 2017, the company has been designing its own IT solutions that help businesses from small to large work efficiently.

For a long time, UniTalk built solutions for clients based on its own hardware. However, servers need to be updated, constantly maintained and there has to be an expert staff to do this. In addition, it was important for the IT company that the services they provide to their customers work without disruption, and therefore they had to be hosted on a reliable platform. If there is downtime, it will cost the company its reputation and customer loyalty.

“We migrated our services to the cloud of one Ukrainian cloud operator”, says Volodymyr Pelykh, UniTalk co-founder and CTO“However, due to the quality of the provided resources, we had to look for another technological partner”.

It was critical for UniTalk that the cloud operator had experience in building cloud infrastructures based on the VMware platform and placed its equipment in several geographically dispersed data centers. This would make it possible to avoid a single point of failure in IT infrastructure.

“We opted for the best. In regard to all criteria, it was GigaCloud”, summarizes Volodymyr Pelykh.

For three years, UniTalk has hosted its services, both internal and for clients, in the GigaCloud cloud — namely, IP telephony and other SaaS solutions.

Different clouds for different tasks

UniTalk uses a VMware-based public cloud in different availability zones: Warsaw and Kyiv. This makes it possible to build a fault-tolerant geographically dispersed infrastructure. On top of that, the IT company has customers within the EU, and the small distance between sites allows reducing the delay between the cloud where the SaaS solution is hosted and the end user.

In addition, UniTalk uses a public cloud with GPU for the development of IT products based on artificial intelligence.

Voice robot and the cloud for AI

One of UniTalk’s products is a voice robot based on artificial intelligence. This is a multifunctional technology for automating communication with customers through phone calls without human intervention. The voice robot makes or receives phone calls, recognizes human speech and engages in dialogue according to a given scenario. In other words, it does all the mundane work and thereby frees up managers’ time for more complex and creative tasks.

To better understand the capabilities of the UniTalk voice robot, let’s give one real-life example. To process a database of 1,844 contacts (to call, have a chat and enter data into CRM), a person will require an average of 39 working days. And the voice robot allowed the client to complete the task in just 3 hours.

The voice robot is based on a cloud with GPU. This GigaCloud cloud is built on NVIDIA Grid technology intended for solving tasks based on artificial intelligence. Virtual GPU technology accelerates resource-demanding server workloads by 10 times (compared to CPU). Processing large volumes of information and voice traffic takes several minutes. This allows the voice operator to make and process from one to several thousand calls simultaneously.

“We are satisfied with the work of GigaCloud technical support and managers who always meet the demands. Currently, we use various cloud infrastructures, which are located in four different data centers of the operator. In order to have a single point of entry and connect different sites into one network, the operator provided us with a dedicated L2 channel. This way, we have a distributed infrastructure, but with a single point of entry and a centralized technical support. It is convenient and reliable to work this way”, says Volodymyr Pelykh, UniTalk co-founder and CTO.