Blue Yonder, the AI company for supply chains, went ahead and announced on May 18, 2026, its Model Training Factory based on NVIDIA Nemotron, to speed up the creation of specialised AI agents when it comes to autonomous supply chains.
Introduced at ICON – Blue Yonder’s annual customer conference – the Model Training Factory is a repeatable system to fine-tune and test highly specialized supply chain models. These models have been taught to carry out high-value tasks with the consistency of supply chain subject-matter specialists, fine-tuned and developed to perform intricate, multi-step supply chain processes, working with human operators and then evaluated to guarantee high-quality outcomes. In the end, through agentic AI, they will allow supply chain processes so as to run autonomously, making decisions throughout warehouse management, demand and supply planning, logistics, and merchandising along with network operations.
Blue Yonder and NVIDIA are teaming up to fine-tune Nemotron open-source models for agent development and to create and launch a platform that combines Nemotron open-source models by NVIDIA and NeMo AI tools with Blue Yonder’s forty years of supply chain decision-making, data and operational experience.
Why it’s important
It is worth noting that supply chain decisioning is incredibly complex, requiring real-time evaluation and collaboration across worldwide distributed teams. It needs extreme accuracy at very low latency throughout thousands of warehouses and lanes as well as stores.
The next generation of AI assistants will assist organisations in analysing what is taking place in the supply chain quicker, with more sophisticated and precise AI. Enterprises are shifting from AI assistants to teams of specialized agents that can comprehend, rationalise, use tools, and perform tasks at machine speed in tandem with human operators. The economics of doing things that, at scale, are moving quickly. As demand for conclusions surges thanks to coding agents, the price of maintaining large frontier models in production continues to climb.
The model training factory based on NVIDIA Nemotron solves these challenges with a hybrid approach – frontier models where a wide range is required and custom supply chain models developed to work with them, providing the speed and accuracy that individual workflows require at just a portion of the cost.
According to Blue Yonder’s CEO, Duncan Angove, “The supply chain has always been an AI domain. Our research into how agentic models perform on real warehouse and planning decision-making is exactly why we know where frontier models hit a wall. Working with NVIDIA, we’re building owned intelligence, not rented intelligence—supply chain models trained on the workflows, telemetry, and decision logic that actually run a warehouse or a planning system. This isn’t a one-off fine-tuned model. It’s a factory, and it produces purpose-built agents at the speed, precision and cost the autonomous supply chain demands.”
Within the model factory
Apparently, Blue Yonder is constructing the Model Training Factory utilising NVIDIA’s agentic AI stack with Nemotron open models as the basis and the NVIDIA NeMo Agent Toolkit to create, evaluate and orchestrate agents. Blue Yonder can adapt the model size to the task with Nemotron’s family of model sizes, from small models optimized for high-speed warehouse decisions to larger models designed for complex multi-step planning.
Each model is taught to be proficient in particular tasks and produce particular results of agentic decision-making and is subject to strict evaluation standards before its deployment and as it evolves over time. Models receive instruction on synthetic data, not on customer data. Blue Yonder is also leveraging NVIDIA AI Enterprise for the Model Training Factory, integrating the microservices, frameworks, and libraries for AI development with high-end GPU orchestration as well as infrastructure management in a single, completely supported, ready-to-use commercial software solution.
Says vice president and general manager, retail and CPG, NVIDIA, Azita Martin, “The next phase of enterprise AI for supply chains requires specialized, affordable and accurate domain-trained agents that can operate within the workflows that run a business. Blue Yonder is leveraging NVIDIA Nemotron, the NVIDIA NeMo Agent Toolkit and NVIDIA AI Enterprise to build a model training factory that fine-tunes models with proprietary supply chain data, enabling them to build agentic AI systems for some of the world’s largest and most complex supply chains.”
First proof points in warehouse
Blue Yonder will deploy the first models towards warehouse management workflows such as WMS allocation shorts, stock exceptions, due-time urgency, and stock across the yard and receiving trailers. These are high-velocity warehouse choices, where accuracy along with quickness impact timely operation, shortages of inventory, and order cycles. Future models will be extended to the larger Blue Yonder solution lineup.
A shift can come apart quite fast in a warehouse. Late trucks, breakdowns in equipment and changing priorities frequently upset a morning plan, requiring continuous reallocation against the clock. A specialised agent can consider hundreds of trade-offs in seconds, whereas an individual typically thinks about a handful, and do it affordably enough to run continually throughout every warehouse, every day.
An advantage that can be repeated
The model factory operationalises expertise and converts it to scalable, reused AI training signals that detect that intelligence in a consistent way regardless of the domain in the supply chain. The benefit of Blue Yonder is the loop itself – workflows, decision logic, telemetry, subject-matter specialists, assessments, and controlled retraining that rivals can’t easily replicate. The first models are anticipated to be deployed into customer production via Blue Yonder Cognitive Solutions later in 2026.
It is well to be noted that NVIDIA is assisting Blue Yonder to go ahead and establish the foundation for a new kind of supply chain AI that is transparent at the model layer, specialised when it comes to the workflow level and built to scale throughout the enterprises that goes on to transport the world’s goods.
