OneRail is using Nvidia-powered AI and optimisation technology to improve how retailers, wholesalers and distributors decide how individual orders should be delivered. The company has launched OmniSTAR, an AI-powered delivery optimisation platform designed to evaluate multiple fulfilment options and select the most cost-effective choice while meeting required delivery service levels.
The platform brings together Nvidia’s cuOpt decision optimisation engine, cuDF data-processing software and GPU-accelerated computing infrastructure with OneRail’s own delivery pricing, operational and performance data.
According to OneRail, the technology can significantly reduce the time required to perform complex delivery calculations. The company says some optimisation workloads that previously required around 20 minutes can now be completed in less than two minutes, while calculations that previously took approximately a week can potentially be completed in about two days.
The reduction in processing time is particularly important for last-mile logistics, where delivery conditions can change quickly. A retailer may need to consider driver availability, traffic, fuel costs, delivery windows, vehicle capacity and pricing before deciding how an order should reach the customer.
What Is OneRail OmniSTAR?
OmniSTAR is designed to help businesses make more intelligent decisions about how orders should be fulfilled and delivered.
Instead of relying on a single delivery method, the platform can evaluate a range of options, including:
- Company-owned delivery fleets
- Third-party couriers
- Parcel carriers
- Delivery service providers
- Other available transportation modes
The system then analyses the available choices and attempts to identify the lowest-cost option that satisfies the required service level.
This is particularly relevant to retailers and distributors that manage thousands or millions of deliveries and need to make fulfilment decisions at scale.
Traditional logistics planning can depend heavily on fixed rules, manual decision-making or previously calculated routes. However, those approaches can become less effective when conditions change during the day.
OmniSTAR is intended to bring optimisation into the operational decision-making process so that businesses can evaluate more options before assigning an order.
Nvidia Technology Behind OmniSTAR
OneRail’s platform combines several Nvidia technologies to accelerate its optimisation and data-processing workloads.
The two key software components are:
Nvidia cuOpt
cuOpt is Nvidia’s GPU-accelerated optimisation technology designed for vehicle routing and other mathematical optimisation problems.
It can consider a variety of constraints when generating solutions, including:
- Vehicle capacity
- Vehicle operating costs
- Travel time
- Distance
- Delivery windows
- Starting locations
- Operational restrictions
- Monetary costs
- Time-based costs
This makes the technology suitable for complex logistics problems where hundreds or thousands of possible combinations may need to be considered.
Nvidia cuDF
OmniSTAR also uses Nvidia cuDF, a GPU-accelerated library for processing tabular data.
Data-processing operations such as filtering, joining and aggregating large datasets can become computationally expensive when organisations are working with millions of records.
GPU acceleration can help process these workloads more quickly, allowing optimisation systems to access and prepare the information they need.
OneRail combines these Nvidia capabilities with its own delivery network information, pricing data and operational models.
How AI Can Improve Last-Mile Delivery Decisions
Last-mile delivery is one of the most expensive and operationally complicated parts of the supply chain.
An order may be fulfilled using different combinations of vehicles, drivers and transportation providers. The cheapest option is not always the best option because businesses must also consider delivery deadlines and customer-service requirements.
For example, a retailer may have several ways to deliver an order:
- Use its own fleet.
- Assign the order to a local courier.
- Use a parcel carrier.
- Use another third-party delivery provider.
Each option can have a different price, estimated arrival time and probability of successful delivery.
OmniSTAR is designed to compare those possibilities before determining which option offers the best balance between cost and service performance.
This can potentially help companies protect margins while maintaining customer delivery expectations.
From AI Prediction to Operational Optimisation
OneRail’s wider AI strategy involves both prediction and optimisation.
These two functions play different roles.
Machine-learning models can predict what is likely to happen, while optimisation systems can use those predictions to determine what should be done.
OneRail says its machine-learning models can estimate factors such as:
- Expected service time
- Risk of late delivery
- Probability of successful first-attempt delivery
- Expected delivery price ranges
These predictions can then become inputs for optimisation models.
The optimisation system can use the available information to determine how a particular delivery should be executed.
This distinction is important in modern logistics because predicting traffic or delivery performance is only one part of the problem. Businesses also need to respond to those predictions with an operational decision.
Dynamic Route Optimisation
Delivery conditions rarely remain constant throughout an operating day.
A driver may become unavailable. A vehicle may break down. A road could be blocked by an accident. Traffic conditions may suddenly change. A high-priority order could also enter the system and require immediate attention.
These changes can make an earlier route plan less efficient.
Nvidia describes cuOpt as a system that can be used for dynamic optimisation by modelling and submitting the updated optimisation problem when operating conditions change.
Examples of events that could trigger a new calculation include:
- Vehicle breakdowns
- Driver absences
- Road closures
- Traffic changes
- New priority orders
- Changes in operating windows
- Capacity changes
OneRail said OmniSTAR can similarly rerun delivery scenarios when variables such as fuel prices, weather and shipping conditions change.
The ability to recalculate more quickly can be valuable because logistics companies may otherwise have to continue operating with a plan that is no longer optimal.
Why Speed Matters in Last-Mile Logistics
The last mile is highly sensitive to time.
A delivery optimisation calculation that takes 20 minutes may already be less useful if the underlying operating conditions have changed significantly during that period.
According to OneRail, OmniSTAR can reduce some 20-minute calculations to less than two minutes.
The company also said workloads that previously required about a week could potentially be reduced to approximately two days.
This faster processing allows optimisation to move closer to real-time operational use.
The underlying idea is simple: the faster a business can evaluate delivery options, the more opportunities it has to make better fulfilment decisions before an order is assigned.
OneRail CEO Bill Catania has emphasised the financial importance of faster last-mile decisions, pointing to the high cost associated with final-mile fulfilment.
How Nvidia cuOpt Searches for Better Routes
Complex logistics optimisation involves an enormous number of possible route combinations.
Testing every possible route would generally be impractical for large-scale real-world delivery operations.
Nvidia says cuOpt does not attempt to exhaustively evaluate every possible route. Instead, the solver generates candidate solutions and repeatedly improves them using GPU-accelerated heuristics.
The objective is to produce a high-quality solution within a specified amount of computation time.
This approach is particularly useful for logistics companies because the goal is often not to prove that a route is mathematically perfect, but to find a highly efficient solution quickly enough to be operationally useful.
OneRail’s Large Delivery Dataset
OneRail combines Nvidia’s optimisation technology with its own delivery data.
The company says its dataset is based on millions of deliveries across a network involving more than 12 million drivers and over 1,000 logistics partners.
The data includes information related to delivery pricing and performance across different transportation modes.
This information can help the system understand how different delivery decisions affect cost and operational performance.
For businesses managing large delivery networks, access to historical delivery information can be particularly valuable because it can reveal patterns that are difficult to identify through manual analysis.
Identifying Hidden Delivery Costs
OneRail says OmniSTAR can also be used to identify delivery rules or configurations that may unnecessarily increase costs.
For example, a business may have an established delivery policy that appears operationally convenient but becomes expensive when applied across thousands of orders.
The system can analyse different scenarios and determine how delivery choices affect profitability at the item level.
This can help companies identify situations where:
- Low-margin products are being transported too far.
- Expensive vehicles are being used for unsuitable deliveries.
- A cheaper delivery provider could meet the same service requirement.
- Existing fulfilment rules are creating unnecessary costs.
- Delivery configurations are negatively affecting margins.
The objective is therefore broader than simply finding faster routes.
The technology is also designed to help companies understand the economics of individual fulfilment decisions.
OmniSTAR in Live Operations
OneRail said OmniSTAR is already being used by selected enterprise customers.
One example cited by the company is US Foods.
According to OneRail, the platform identified delivery configurations that were negatively affecting margins. One example involved low-margin products being transported long distances using higher-cost equipment.
US Foods subsequently used those findings to modify pricing and restructure some delivery patterns, according to OneRail.
The example illustrates how optimisation technology can potentially influence decisions beyond routing.
Instead of simply telling a driver which road to take, a system can help businesses examine whether the overall delivery configuration makes economic sense.
Reported $40 Million Savings
OneRail has also reported significant financial benefits for another large customer.
The company told CNBC that an unnamed large tyre distributor using the platform achieved approximately $40 million in run-rate savings over three years.
The customer was not identified publicly, and the savings figure was provided by OneRail.
Such figures should therefore be viewed as a company-reported customer result rather than an independently verified industry-wide benchmark.
Nevertheless, the claim illustrates the type of financial opportunity OneRail believes AI-powered fulfilment optimisation can create for large logistics operations.
OmniSTAR and $6 Billion in GMV
OneRail also told CNBC that it expects OmniSTAR to exceed $6 billion in gross merchandise volume during the fourth quarter of 2026.
GMV refers to the total value of goods transacted through a platform or associated commerce activity and is not the same as revenue.
The projected figure highlights the scale at which OneRail expects the technology to operate as more enterprise customers adopt the platform.
Three Years of OneRail and Nvidia Collaboration
According to CNBC, OneRail and Nvidia worked together on the project for approximately three years before the launch of OmniSTAR.
The collaboration included direct engagement with Nvidia’s cuOpt engineering team on last-mile delivery and large-scale logistics optimisation.
OneRail also participated in the Nvidia Inception programme, which supports startups working on technology and innovation.
The long development period reflects the complexity of applying mathematical optimisation and GPU computing to real-world delivery operations.
Logistics optimisation is not simply a matter of calculating the shortest distance between two locations. Real systems need to consider costs, capacities, operating restrictions, delivery windows and changing conditions.
Why Retailers Are Turning to AI for Delivery
Retailers increasingly need to balance two competing priorities:
Lower delivery costs and better customer service.
Consumers expect faster and more reliable deliveries, while businesses face pressure to control fulfilment expenses.
This creates a difficult optimisation problem.
A retailer cannot necessarily choose the cheapest delivery option if it causes an order to arrive late. At the same time, consistently selecting premium delivery services can significantly reduce margins.
AI and mathematical optimisation can help businesses evaluate these trade-offs at a much larger scale than manual planning.
The potential benefit is particularly significant for businesses handling large order volumes.
OneRail and FedEx SameDay Local
OneRail’s relationship with major logistics companies extends beyond OmniSTAR.
In March, FedEx launched FedEx SameDay Local in collaboration with OneRail, connecting customers to a national network of more than 1,000 delivery providers.
The collaboration demonstrates the growing role of technology platforms in coordinating fragmented last-mile delivery networks.
Rather than relying entirely on a single delivery fleet, platforms can connect businesses with multiple delivery providers and help determine how orders should be fulfilled.
The Future of AI-Powered Last-Mile Delivery
The logistics industry is moving toward increasingly dynamic decision-making.
Traditional delivery systems often depend on predefined rules and schedules. AI-powered optimisation introduces the possibility of continuously evaluating changing operational conditions.
Future delivery optimisation systems are likely to combine several technologies, including:
- Machine learning
- Predictive analytics
- Mathematical optimisation
- GPU computing
- Real-time operational data
- Dynamic routing
- Automated fulfilment decisions
The combination of prediction and optimisation is particularly important.
Prediction can estimate what is likely to happen, while optimisation can determine the best response.
For a delivery business, this could mean predicting traffic and delivery delays and then automatically selecting another fulfilment option before the order is assigned.
Conclusion
OneRail’s OmniSTAR represents an effort to bring GPU-accelerated AI and mathematical optimisation into the day-to-day economics of last-mile delivery.
By combining Nvidia cuOpt, cuDF and accelerated computing infrastructure with OneRail’s delivery pricing and performance data, the company says it can evaluate delivery scenarios substantially faster than its previous approach.
The platform is designed to compare owned fleets, couriers, parcel carriers and other transportation options while considering both cost and required service levels.
Its potential value goes beyond route planning. OneRail says the system can identify expensive fulfilment configurations, assess item-level profitability and help companies adjust delivery strategies as conditions change.
The company’s reported customer examples, including the US Foods case and the claimed $40 million run-rate savings for an unnamed tyre distributor, point to the broader commercial opportunity. However, these figures are company-reported and should not be interpreted as independently verified benchmarks.
As retailers and distributors continue to face rising pressure around delivery costs and customer expectations, faster optimisation could become an increasingly important part of supply-chain technology.
With GPU computing making complex calculations faster and AI models providing increasingly detailed predictions, the next generation of last-mile logistics may rely less on static delivery rules and more on real-time, data-driven fulfilment decisions.
Note: Claims regarding processing-time improvements, customer savings, delivery-network scale and projected GMV are based on statements attributed to OneRail and its reported partners in the supplied source material.
Discover more from AiTechtonic - AI & Informative News
Subscribe to get the latest posts sent to your email.