From March 16–19, 2026, the world’s most influential AI conference, NVIDIA GTC, took place in San Jose, alongside a global online livestream.

As a global leader in AI retail technology, SandStar announced that its Founder and CEO, Yili Wu, was invited to speak at GTC alongside Takashi Tokita, Deputy General Manager of Lawson China, and Ekaraj Panjavinin, Chief Digital Officer of Thailand’s True Corporation.

Their joint session, titled “Reimagining Store Operations at Scale to Accelerate Retail ROI,” explored how AI can be systematically deployed across large retail networks to drive operational efficiency and sustainable returns.

Key Session Focus: Easy‑to‑Scale AI for Retail with 3–6 Month Payback

At the conference, Yili Wu draw on SandStar’s ten years of hands‑on experience in AI‑powered retail, presenting real‑world case studies from AI store and AI vending machine. The session traced how retailer demand for AI had evolved over the past decade and outlined a clear path from AI deployment to measurable business value.

At the technology level, Wu shared SandStar’s practical experience in building stable, scalable AI systems in complex retail environments—leveraging NVIDIA Jetson–based edge computing, NVAIE, and Cosmos model capabilities—to enable continuous algorithm optimization and large‑scale global deployment.

Representatives from Lawson and True Corporation shared firsthand implementation experiences from AI‑powered store operations, covering core scenarios such as traffic analytics, demand‑driven ordering and replenishment, merchandising, and marketing decision‑making. These case studies illustrated how AI can be embedded directly into frontline workflows to support store staff, managers, and regional operations teams—achieving payback within three to six months and delivering long‑term performance advantages that are scalable and repeatable across markets.

Stepping Onto the GTC Stage: A Decade of Commitment and Value Creation

As the global barometer for AI innovation, NVIDIA GTC brought together leading innovators from across industries. SandStar’s invitation to speak at GTC reflected its decade‑long commitment to AI retail and its unwavering user‑centric philosophy.

A Decade of Deep Expertise Serving an Evolving Industry

In AI vending, AI has progressed from basic product recognition to comprehensive operational support—enabling location‑specific assortment planning, optimized replenishment routes and timing based on sales peaks, and improved loss prevention.

At the store level, AI has driven the evolution from rigid, uniform SOPs to dynamically optimized, store‑specific operating models, adapted to store size, location, competitive context, and customer profiles—helping stores respond faster to local demand without increasing management complexity.

At the workforce level, AI’s role has shifted from being perceived as a monitoring tool to becoming an everyday operational assistant, offering real‑time guidance that allows staff to focus on higher‑value customer interactions while improving decision accuracy and reducing missed opportunities.

Building Core Capabilities Through Sustained Investment

In 2017, when SandStar chose to rely on AI to create value for retail, few believed that AI could reliably recognize products in open, highly complex physical stores—let alone generate real‑time insights and operational decisions.

Over the past decade, SandStar has invested millions of dollars, building a technology team of over 200 people and continuously advancing across hardware, algorithms, and software.

From initially deploying a single NVIDIA 1080 Ti GPU as an edge computing starting point, SandStar has completed seven generations of edge computing and camera upgrades, reducing hardware costs by 70% while fully supporting real‑time AI recognition at the edge. On the data side, SandStar has accumulated datasets covering more than 1.5 million SKUs globally and over 10 million hours of retail‑scene data. To address industry challenges such as high labeling costs and long learning cycles, SandStar developed a proprietary algorithm feedback loop, shortening new‑SKU learning time from 7–10 days to approximately 10 minutes, while maintaining industry‑leading recognition accuracy.

To enable low-cost and high-efficiency AI deployment, SandStar has established a mature edge-cloud collaborative architecture. Validated in complex environments over years of operation, the system delivers stable edge execution with self‑healing mechanisms and local caching, ensuring data security and business continuity under extreme conditions. Built on a unified, flexible, stateless architecture, the platform supports rapid deployment and large‑scale operations across cloud, edge, and local environments. Through continuous algorithm feedback from the edge, cloud models are incrementally optimized, enabling ongoing improvements in recognition and analytics. Resource allocation is dynamically adjusted by task, achieving an optimal balance between performance and cost.

To simplify store operations and reduce potential loss, SandStar has integrated NVIDIA’s Vision System Software (VSS) foundation vision model. With traditional computer vision methods, operational updates—such as changes to employee uniforms—could take up to a month to deploy. With NVIDIA VSS, the same update can now be implemented in as little as three days. More importantly, traditional computer vision systems struggled to monitor long, multi‑step workflows, such as suspected shoplifting. NVIDIA VSS understands shoplifting as a complex sequence of actions rather than a single event, enabling continuous tracking and interpretation of the full process, even when it spans several minutes.

In the era of Agentic AI, SandStar has introduced AI Agents to create a dedicated “Retail Operations Expert” for each store. These agents are trained on tens of thousands of hours of operational experience from more than 300 top retail managers worldwide and continue to evolve based on daily, real‑world store data. For retailers, this is equivalent to having hundreds of specialized operations experts working simultaneously in each store—covering assortment planning, ordering, replenishment cadence, merchandising optimization, promotions, workforce scheduling, and task execution. Through SandStar’s ChatStore, operators can interact with these AI agents in real time, receiving actionable insights and executable recommendations tailored to current business conditions—making professional retail expertise truly scalable.

User‑Centric by Design: Creating Value Through Deep Operational Engagement

Unlike approaches that prioritize technology over real‑world implementation, SandStar’s core strength lies in deep expertise in both AI and retail operations. Team members are required to work closely with customers on the frontline, embedding AI into business processes rather than delivering technology in isolation. 

In its collaboration with Lawson, SandStar’s AI Store Solutions team stationed itself in frontline stores for over a month, gaining a detailed understanding of site selection, space layout, product structure, merchandising logic, and operational rhythm—ensuring AI capabilities are truly usable, effective, and value‑creating.

Guided by this principle, every technological innovation at SandStar is designed to address genuine retail pain points: AI‑driven ordering and replenishment to balance overstock and out‑of‑stock risks; on‑device privacy protections such as real‑time facial masking to address compliance requirements; and flexible, scalable deployment architectures that allow 95% of existing cameras to be reused, significantly lowering the threshold and cost of AI transformation.

Deployed in 30 Countries: Advancing the Globalization of Retail AI

Over the past decade, SandStar has focused exclusively on AI applications in retail, expanding its presence to 30 countries and regions, serving more than 30 Fortune 500 companies, and supporting tens of thousands of growing retailers. To date, SandStar has delivered AI capabilities to nearly 500,000 stores and retail terminals, with systems operating long‑term across diverse geographies and formats.

In its global expansion, SandStar prioritizes long‑term sustainability—adapting to local infrastructure, regulatory requirements, consumer behavior, and operational complexity—ensuring AI solutions can be deployed securely, reliably, and at scale.

Through years of cross‑regional execution, SandStar has developed a repeatable methodology for retail AI adoption, from scenario selection and system deployment to continuous optimization—offering the industry a practical blueprint for long‑term value creation.

Join NVIDIA GTC 2026 and Explore the Next Decade of AI Retail

As global retail enters a pivotal phase of digital and intelligent transformation, IDC forecasts that the global retail AI market will reach USD 45 billion by 2025. Whether you are a retail executive seeking scalable growth, a technology professional or investor exploring real‑world AI adoption, or simply interested in edge intelligence and AI retail, join NVIDIA GTC 2026 (March 16–19 | San Jose & Online) and attend the session by SandStar Founder & CEO Yili Wu.

 Session link: https://www.nvidia.com/gtc/session-catalog/sessions/gtc26-s81675/