Ringover is a prominent cloud communication provider, delivering advanced VoIP, video conferencing, and AI-powered tools to businesses. Known for reliability and seamless CRM integration, Ringover handles over 2.5 million calls daily. Their AI tools enhance services like speech-to-text and emotion detection, allowing companies worldwide to streamline communications, improve customer interactions, and maintain high security and compliance standards.
How DataPacket powers Ringover’s AI and telephony solutions
As a leader in cloud-based communication, Ringover required a reliable infrastructure partner to support its rapid growth, particularly for its AI-driven tools and high-performance telephony systems. Partnering with DataPacket allowed Ringover to scale effortlessly while maintaining peak performance.
Scalability and performance
Handling millions of calls daily, Ringover relies on AI models for natural language processing (NLP), automatic speech recognition (ASR), and large language models (LLMs). These requirements demanded scalable, high-performance infrastructure in multiple locations, ensuring low latency and compliance with data regulations.
As a company with consistent growth, supporting clients in demanding and competitive business environments, we were looking for a reliable partner, which was able to support our growth. DataPacket has proven very efficient in this area and has offered the scalable and high-performing infrastructure we needed to serve our clients in various locations worldwide.
Custom high-performance servers with advanced GPU capabilities
To meet Ringover’s AI and telephony needs, DataPacket provided specialized infrastructure, including Nvidia-powered A40 and L40S servers. These high-performance GPUs are essential for accelerating AI tasks in Ringover’s Empower and Babel modules. The A40 GPUs are designed for data center workloads, providing the power needed for AI models, including emotion detection, named entity recognition (NER), and large language models (LLMs).
DataPacket’s servers being located worldwide is a true strength. This allows us to expand our presence quickly and efficiently, especially when entering new markets outside of Western Europe. This geographical flexibility is crucial to supporting our global growth.
Additionally, this diversity in locations has greatly facilitated our entry into the GPU server rental space for artificial intelligence. We use them particularly for applications such as emotion detection, named entity recognition (NER), large language models (LLMs), and transcription.
Server configuration and AI processing
Ringover’s AI models rely heavily on GPUs for large-scale NLP tasks. Key tasks include:
- Emotion and mood detection: GPU-intensive processes analyzing voice data to gauge sentiment.
- Named Entity Recognition (NER): Extracting key information (names, phone numbers, etc.) from transcripts using deep learning models.
- Automatic Speech Recognition (ASR): Converting speech to text with high accuracy across languages.
- Text-To-Speech (TTS): GPU-enhanced for faster, natural-sounding speech generation with minimal latency.
- Speaker Diarization: Identifying and separating speakers in audio segments, essential as Ringover processes up to 2.5 million calls daily.
CPU-Driven processing
While GPUs manage AI tasks, CPUs are optimized for:
- Waveform generation: Creating sound waves from processed text in TTS applications.
- Audio resampling and segmentation: Splitting and assembling transcribed audio segments.
- Waveform analysis: Pinpointing key moments in audio for storage and retrieval.
DataPacket’s servers also provide:
- High-bandwidth, low-latency connectivity: Vital for real-time communication tasks like live NLP processing.
- 10 Gbps uplinks: Ensuring low-latency, high-speed connections between AI models and users.
Containerized deployment for flexibility
Ringover deploys its applications in containers on DataPacket’s servers, offering flexibility and quick scalability. Containerization enables smooth application transfers across servers without compatibility issues, facilitating fast deployments, minimizing downtime, and allowing Ringover to scale based on real-time demand.
Data compliance and global presence
With a growing global customer base, Ringover must comply with regional data retention laws while minimizing latency. DataPacket’s global presence, with servers in regions like the USA and EMEA, ensured compliance and reduced latency by keeping data closer to users, improving the communication experience.
Custom solutions and on-demand scalability
DataPacket’s infrastructure allows Ringover to scale quickly. Servers are provided on demand, giving Ringover the flexibility to scale resources as needed. This adaptability is crucial for handling peak traffic and launching new features, like AI enhancements in the Empower product.
By utilizing Nvidia-powered A40 and L40S servers, Ringover can fine-tune and rapidly deploy AI models across its global network, improving user experience without compromising performance.
Low-latency connectivity for multichannel communication
Low-latency connectivity for multichannel communication is crucial to Ringover’s stable, scalable videoconferencing. To achieve this, Ringover leverages DataPacket’s low-latency network to support seamless communication across its global infrastructure.
In Paris, Ringover’s largest location by server count, DataPacket provides direct interconnections with France’s top three ISPs: Orange, SFR, and Bouygues Telecom. This results in about 75% of destinations across France being reached with optimal efficiency, often achieving single-digit latency.
At all 57 of our locations, we prioritize direct interconnections with local ISPs. This strategy reduces latency by avoiding inefficient, congested, or distant routes, delivering consistent low-latency performance that benefits everyone.
For remaining destinations, we rely on a fully redundant peering mix, including Tier 1 transit carriers, private interconnections, and local IX points. Underperforming routes are continuously monitored and adjusted, which is essential in today’s ever-changing internet landscape.
Seamless scalability and enhanced AI capabilities
Since partnering with DataPacket, Ringover has grown consistently, expanding from 12 servers in 2020 to over 100 today. DataPacket’s vast network and rapid server deployments have allowed Ringover to meet rising customer demand and maintain the low latency crucial for its global communication services.
Proactive support and downtime mitigation
Ringover benefits from DataPacket’s 24/7 support through a dedicated Slack channel, ensuring swift resolution of issues. During two downtime incidents in Paris, DataPacket’s prompt intervention minimized disruptions, reinforcing their reliability and commitment to uptime.
Partnership for growth
DataPacket’s tailored infrastructure solutions have enabled Ringover to meet the growing demands of its AI models while scaling globally. With ongoing support and flexible infrastructure, Ringover is well-positioned for continued expansion in the cloud communication space.
We are extremely satisfied with the server rental service provided by DataPacket, which we consider to be ultra-premium. Their support is of the highest quality, with outstanding responsiveness, making a significant difference in our day-to-day collaboration. Having privileged contacts and tailored solutions demonstrates their commitment to truly understanding our specific needs and adapting flexibly.
This level of service and personalization creates a real sense of confidence. We really feel like we are working with a partner who cares as much about our success as we do, which makes working with DataPacket both enjoyable and motivating. The team’s availability and expertise add a human touch to our partnership that is simply irreplaceable.