OneRail uses Nvidia AI for real-time last-mile delivery optimisation

Details

Last-mile delivery optimization has traditionally run in batches, taking 20 minutes or more to evaluate carrier options across owned fleets, couriers, and parcel services. OneRail's new OmniSTAR platform changes that by integrating Nvidia cuOpt — an open-source GPU-accelerated routing library — and Nvidia cuDF for data processing, delivering up to 10x faster calculations. That speed difference is operationally significant: OmniSTAR can now run within live delivery workflows, comparing multiple fulfillment modes and selecting the lowest-cost option that meets a required service level before an individual order is assigned. Calculations that previously required a week can now be completed in approximately two days for full network re-optimization. OneRail's ML models feed into this system by predicting service time, lateness risk, and first-attempt delivery probability. CEO Tony Catania put the business case plainly in a CNBC interview: 'If you don't have the ability to make lightning-fast decisions, you're giving up margin.' The platform targets retailers, wholesalers, and distributors. Notably, Nvidia cuOpt is open-source and documented for vehicle costs, capacities, travel times, and operating windows — meaning competitors can integrate the same GPU-accelerated routing engine independently of OneRail.