Jia Li

AI Summary

→ WHAT IT COVERS Jia Li, cofounder and chief AI officer of LiveX AI, explains how full-size holographic AI agents transform fan experiences at Super Bowl 2026 and retail environments. The company deploys human-like holograms running on NVIDIA RTX 6000 GPUs to provide real-time wayfinding, customer service, and personalized interactions across 20 activations during Super Bowl week. → KEY INSIGHTS - **Physical retail gap:** 80% of US shopping still occurs in physical stores, not online, creating a massive opportunity for AI agents to bridge the digital-physical divide. Retailers lack data on in-store customer needs and behavior, while brands cannot provide the same intelligent, personalized assistance available through digital chatbots. Full-size holographic agents address this by offering real-time inventory checks, product recommendations, and seamless checkout experiences without requiring customers to wait for human staff or resort to mobile searches. - **Technical performance requirements:** Mission-critical AI agents require both efficiency and accuracy for real-time interactions. LiveX AI achieves six times faster average token speed using NVIDIA NIM microservices compared to traditional inference frameworks. This speed improvement enables multi-step AI agent workflows where each step involves language or vision models, allowing six interaction steps in the time competitors complete one. The system runs on-premise using two NVIDIA RTX 6000 GPUs with Blackwell architecture for simultaneous visual rendering and large model inference. - **Super Bowl deployment scale:** LiveX AI operates over 20 holographic AI agent activations across Super Bowl week 2026, spanning airports, fan zones, city halls, and concert venues. The agents, named Lyra, handle wayfinding, event information, emergency protocols, and crowd management for millions of fans. The system scales using both on-premise RTX workstations and cloud deployments to handle peak volumes, then scales down during lower traffic periods. This represents the first deployment of 4K high-quality, full-size holographic agents at this scale. - **Production stack architecture:** The system combines multiple model types including custom-trained models and third-party language models for languages the team cannot evaluate internally. LiveX AI uses NVIDIA Triton, TensorRT, NeMo, and Nematron across the stack, with models deployed on Kubernetes at Google Cloud for auto-scaling capabilities. The team optimizes at the kernel level using CUDA to maximize computational efficiency. On-premise deployment proves essential for crowded indoor events like NRF with 40,000 attendees where internet connectivity becomes unreliable. - **Athlete-fan relationship evolution:** Human-like AI agents enable celebrities and athletes to maintain authentic connections with fans at scale without physical presence. Sports teams prioritize fan experience above all else, seeking ways to provide lifetime memorable moments through interactions with players and team history. AI agents can represent individual athletes, answer questions about teams and leagues, facilitate celebration selfies with virtual characters, and create personalized experiences that previously required direct human interaction, fundamentally changing how fans engage with sports properties. → NOTABLE MOMENT Li describes how retail customers already pull out phones to use chatbots when they cannot find products in stores, but those chatbots do not represent the brand, lack current inventory data, and often provide outdated information. This behavior reveals customers already prefer AI assistance over waiting for human staff, creating immediate demand for brand-controlled holographic agents with accurate real-time information. 💼 SPONSORS [{"name": "NVIDIA GTC", "url": "nvidia.com/gtc"}] 🏷️ Holographic AI, Retail Technology, Sports Fan Engagement, NVIDIA RTX, Physical AI Agents