Customer Service Automation · Real-time Analytics
Singtel needed to make its 5G enterprise offering compelling for AI-heavy customers, but real-time video analytics, AR/VR, and chatbot workloads required low-latency edge compute, shared GPU capacity, and production-ready AI software that ordinary network infrastructure could not provide.
Reported impact
Singtel created a repeatable edge-AI platform for 5G customers, shortened experimentation with chatbot and video workloads, and built a production path for low-latency enterprise services that can be replicated by regional telcos as Paragon expands globally.
Singtel built its multi-access edge compute environment around NVIDIA AI Enterprise so enterprise customers can run AI and graphics-heavy applications directly on the 5G edge instead of shipping them back to centralized cloud regions. The MEC stack combines T4 and A40 GPUs with GPU virtualization so multiple tenants can share accelerated infrastructure, while the Singtel Paragon platform provides the orchestration layer for deploying and managing those services. On top of that base, Singtel uses tools such as Riva and NeMo to prototype conversational experiences and exposes frameworks like DeepStream and PyTorch for customer video analytics workloads. That architecture gives customers one path from pilot to production across chatbots, edge video processing, and immersive applications, while letting Singtel extend validated 5G plus AI patterns beyond its home market through Paragon's global rollout.
Enterprise applications are deployed through the Paragon platform onto MEC infrastructure backed by T4 and A40 GPUs. AI Enterprise virtualizes and manages the accelerated runtime, while services such as chatbots and video analytics process data at the 5G edge and return results locally with low latency.
For example, customers are already leveraging NVIDIA DeepStream, PyTorch libraries-included in NVIDIA AI Enterprise-for video analytics applications.
Singtel's MEC platform includes NVIDIA T4 Tensor Core and A40 GPUs.
The MEC infrastructure leverages NVIDIA accelerated compute and NVIDIA AI Enterprise to support AI and graphics-intensive use cases.
To handle multiple concurrent users, Singtel turned to GPU virtualization with NVIDIA AI Enterprise to effectively share GPU resources across a broad range of AI 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
OpenAI Sample Jul 10, 2026 Deutsche Telekom Deutsche Telekom needed to redesign customer care, network operations, and employee workflows at telecom scale, rather than adding isolated AI tools to legacy processes. 50,000+ monthly active users of ChatGPT and API tooling AI-powered customer service · ChatGPT Enterprise · In-call assistants 
Google Cloud Sample Archived · Jun 27, 2026 TELUS TELUS needed to break down channel silos, analyze far more than a tiny fraction of its support interactions, and automate customer-service resolution without exposing sensitive data or forcing customers and agents to… $53.9M in annual operational cost savings BigQuery · Gemini for customer experience · Gemini platform 
NVIDIA Sample Feb 2, 2026 Lowe’s Lowe’s needed to close product-expertise gaps for project-based shoppers and store associates, gain better visibility inside large stores, and modernize slow, expensive merchandising and planogram processes across its… 1,750+ stores with digital twins updated throughout the day Dell PowerEdge XR Series servers · Hyper3D.ai’s Rodin · Mylow 