Document Processing
RingCentral needed a transcription platform that could handle hundreds of thousands of concurrent meeting users, different accents, specialized vocabulary, and noisy audio while still producing real-time results.
Reported impact
RingCentral improved transcription accuracy by more than 10 percent, supported over a billion minutes of meetings for 200,000 concurrent users, and raised the quality of downstream summarization and sentiment-analysis tasks.
RingCentral modernized its transcription stack by taking pretrained NVIDIA speech models and adapting them with proprietary meeting data in NVIDIA NeMo, then deploying the resulting speech-recognition service with NVIDIA Riva for production use. NVIDIA DGX A100 systems provide the accelerated environment for training and experimentation, while NVIDIA Triton Inference Server supports scalable inference across large numbers of simultaneous audio streams. The end-to-end pipeline is built to handle diverse accents, domain-specific jargon, and noisy meeting conditions so transcripts arrive in real time and remain useful for downstream tasks. Those transcripts feed collaboration features such as live captioning, meeting summaries, and sentiment-oriented analysis, allowing RingCentral to support more than a billion meeting minutes without sacrificing speed or accuracy.
Meeting audio is processed by NeMo-tuned ASR models trained on DGX A100 systems, Riva serves live transcription in production, and the resulting transcripts flow into captioning, summarization, and sentiment-analysis features.
Accuracy increased by more than 10 percent
RingCentral fine-tuned NVIDIA's state-of-the-art, pretrained speech recognition models on proprietary custom data with NVIDIA NeMo.
RingCentral, a leading provider of unified-communications-as-a-service (UCaaS) solutions, transcribes over a billion minutes of meetings for 200,000 concurrent users on their platform.
The models were deployed in production using NVIDIA Riva-a GPU-accelerated SDK for deploying world-class AI-based speech applications.
Review how this reference was scored and trace each normalized claim back to its documented source.
Scoring methodology
Customer Reference Value measures usefulness as a customer reference. Evidence Confidence measures the quality and traceability of available evidence, not independent verification.
Traceability
Field coverage shows which facts each source supports. Dates are shown only when documented by the source.
Related
NVIDIA Sample Sep 22, 2026 Sluicebox Sluicebox needed to turn fragmented bills of materials, supplier records, PDFs, CAD files, and emissions inventories into timely product carbon assessments, but manual life-cycle-analysis work was slow, expensive, and… 51-80% lower inference costs Lucy · NVIDIA Nemotron 3 Ultra · carbon graph NVIDIA Sample Sep 1, 2026 Onetera Onetera needed to help more than 90 municipalities turn fragmented regulations, zoning maps, permit rules, and legacy documents into fast resident-facing services without imposing heavy IT overhead or expensive manual… 1.7x increase in data throughput GPT OSS 20B · Government Intelligence Agent · LangGraph NVIDIA Sample Jun 26, 2026 MISA MISA needed to move from expensive, externally hosted LLM APIs to a self-hosted Vietnamese AI stack that could protect sensitive data, reduce latency and operating cost, and support many concurrent public-service and… $200,000 in monthly cost savings MISA AVA · NVIDIA A40 GPUs · NVIDIA DGX H200 