At its air hubs, a global express carrier uses FreightLM to read air waybills, commercial invoices and customs paperwork the moment they arrive, so shipments clear faster and teams focus on exceptions instead of data entry.
You move global trade forward.
FreightLM powers the paperwork.
Every shipment generates information,
but most organizations still process it by hand.
FreightLM reads, understands and acts
on freight paperwork automatically,
so your teams spend less time handling
documents and more time moving the world.
Performance
Small model. Serious results.
The challenge
Documents are the freight industry’s
last major bottleneck.
Global supply chains are faster, more connected, and more data-driven than ever before.
Yet the information that powers them remains trapped inside bills of lading, customs declarations, invoices, packing lists, and hundreds of carrier-specific document formats.
A top-25 forwarder handles 130m+ documents per year.
sheets of paper per shipment on average
stakeholders per shipment on average
The solution
FreightLM turns
documents into decisions.
Instead of relying on thousands of templates, FreightLM uses a specialised vision-language model trained specifically on freight documentation.

Success stories
Already at work.
By land, sea and air.
Six months after deployment, a global freight forwarder has automated the processing of millions of freight documents, cutting processing time by 70% while improving data quality across operations.
On its ocean trade lanes, a container carrier uses FreightLM to split mixed shipping bundles into bills of lading, packing lists and certificates of origin, and to read each one with the source shown beside every value.
How it works
Capture. Understand. Act.
Inbuilt industry intelligence
Shipping documents, emails, contracts, operational records. FreightLM knows what to expect, and understands the unexpected.
Find what matters
Extract, classify and validate critical information with minimal human intervention.
Act before bottlenecks
Make operational decisions, with zero change in upstream and downstream processing.
Architecture
Compact by design.
Sovereign by default.
A 3-billion-parameter vision-language model.
Fine-tuned on freight documents, the model identifies the document type, finds the boundaries inside a multi-page bundle, then extracts the key fields. One inference pass, straight from the page image.
Production-scale efficiency.
Unlike large foundation models that demand extensive infrastructure, FreightLM is deliberately compact. Trained and evaluated on 2–4 GPUs in pilot, scaled to 8–10 GPU clusters in production. No hyperscale infrastructure required.
Plugs into the systems you already run.
CargoWise, SAP TM, ERP platforms and in-house systems. Validated outputs are delivered directly into existing workflows: no operational change, no platform replacement, at a fraction of the cost per document.
Freight processing that never leaves your network.
The platform is freight-native, tokenless, on-premise and vendor-neutral – four qualities rarely found together. It’s small enough to run on infrastructure you already own, yet specialised enough to solve freight workflows at scale.