Multimodal Freight Forwarding Platform with Intelligent Workflow Automation
We built a bilingual multimodal freight forwarding and logistics platform with an applied API and RPA operations layer from scratch, modernizing fragmented and email-based forwarding. The logistics platform runs document intelligence with automated customs data extraction, predictive ETA with multi-carrier status tracking, an AI rate engine, invoice reconciliation, and HS code classification, giving HJM a real-time, connected operational workflow across road, air, sea, and rail.
- Bilingual NL/EN Brand Site
- AI Rate Engine
- Document Intelligence
- Predictive ETA
- HS Code Classification
- Client Tracking Portal
- Internal Ops Console
About the Project & Business Goal
HJM Logistics, a Hae Joong Group multimodal freight forwarder in the Netherlands, has strategically partnered with us to modernize its digital presence & operations. The business goal was simply to eliminate the slow, email-driven freight processes, improve the quoting process, automate documentation, & provide customers with greater shipment visibility. Developed a bilingual platform with an applied AI operations layer that features intelligent freight quoting, document intelligence, predictive ETA, HS code classification, a client tracking portal, and an internal operations console, establishing a faster and more connected freight forwarding ecosystem.
Key Outcomes Delivered
Reduced Quote Turnaround to Seconds
Our AI rate engine provides indicative quotes in under 2 seconds, thus replacing manual email-based pricing with instant customer responses.
Eliminated Manual Document Entry
Our intelligent document system automatically parses Bills of Lading & custom documents, replacing manual re-keying with quick verification system.
Delivered Predictable Shipment ETAs
Our predictive ETA model generally combines historical & live shipment signals to deliver more reliable arrival forecasts that help in operations.
Numbers that Define the Platform
We developed HJM logistics with an applied AI operations layer that helps in automating quoting, document processing, & also in shipment intelligence that guides teams to replace the manual freight operations with faster, data-driven approaches across each & every stage of the overall process.
Faster First-Pass Quoting
Our AI rate engine helps to reduce indicative quote turnaround from hours of manual setup to under two seconds for faster customer responses.
Fields Auto-Extracted
The document smart intelligence is able to extract shipment datasets from Bill of Lading, invoices, packing lists, & custom documents with minimal review.
Production AI Workloads
Deployed 4 production AI capabilities: rate engine, document intelligence, predictive ETA, & HS code classification supporting daily freight processes.
Less Manual Data Entry
More Accurate ETAs
AI Decisions Auditable
Faster Document Handling
Engineering & Product Challenges We Solved
Freight forwarding is a particular domain where small mistakes can have major consequences. Any wrong HS code delays customs, missed milestones risk contracts, & unreliable ETAs weaken procurement trust. Enabling AI in the workflow makes operations much more seamless and reliable, not more complex.
Pricing a Multimodal Move in Real Time
The freight pricings generally depends on lanes, modes, containers, surcharges, seasonality & market rates, thus requiring a rate engine system that is able to match experienced forwarder decisions.
Every B/L Is a Different PDF
There are all different formats in Bills of Lading, invoices, packing lists, & certificates. For maintaining a great unified process, we need to have a structured document intelligence other than the traditional OCR process.
Carrier ETAs Are Overly Optimistic
The carrier ETAs ignore prices in port congestion, weather, customs delays & vessel history. As a result, this requires predictive ETA models that tend to consistently outperform published arrival estimates.
Customs Codes, Customs Officers Will Accept
HS codes determine duties, restrictions & clearance speed, thus creating a need for accurate AI suggestions that HJM logistics operators could easily verify prior to customs submissions in the work process.
AI Decisions a Forwarder Can Defend
Each AI-generated quote, document field, ETA, & custom code needs to have audit trails, confidence scores & human oversight instead of black-box decisions. Every output also had to remain fully traceable.
Dutch & English With Equal Editorial Quality
Our platform requires native quality Dutch & English experiences across the website, rate engine system, client portal, quotes, and customer communication workflows that help ensure equal usability.
Solutions We Delivered for HJM Logistics
Every decision focused on one specific outcome; i.e., reducing manual operational efforts while enabling HJM’s operations team to prioritize freight planning, customs expertise & customer service as the platform automated core logistics processes.
We developed a machine learning rate engine system that is trained on historical lane rates, container prices, fuel charges, seasonality & live market data. Thus, indicative quotes are generated within seconds, labelled as AI-generated, and made auditable with pricing factors for better validation.
Developed a 2-stage document pipeline system in enterprise OCR that captures document content & LLM extracts structured shipment datasets from Bills of Lading, invoices, & customs documents. Associated with confidence scoring, helps in automatically routing extraction without any issues.
Built a predictive ETA model utilizing historical shipment performance along with live port congestion, vessel tracking, weather status & lane level patterns. Along with the confidence scores, ETA predictions actually refreshed, giving better output to the customers on arrival forecasts than any rough estimates.
We developed an HS classification assistant system that actually combines a fine-tuned LLM aligned with custom knowledge. The product descriptions generate ranked HS code suggestions, confidence scores, duty implications, document requirements, improving the future recommendations.
We established an internal operations console system that helps in centralizing AI-generated quotes, parsed documents, HS code suggestions & shipment milestones. Along with the comprehensive audit logs, there is a capture of every AI input, model version, & other data for great traceability.
Modelled EU customs workflows, transit documentation, AEO status, dangerous goods classifications, air freight references, and also Dutch VAT managing directly into the shipment data model. Additionally, metadata of compliance associated with shipments ensures accuracy.
Designed for Shippers. Optimised for Operations.
We designed every workflow around two distinct user groups, customers requesting freight services and HJM’s operations team managing shipments. The result was a bilingual experience with role-specific interfaces that simplified freight quoting, shipment tracking, document review, and AI-assisted logistics operations without fragmenting the platform.
Production AI Workloads We Deployed
There are 4 production AI workloads forms the intelligence layer of a platform comprising 9 integrated capabilities, including 4 customer-facing experiences and 1 internal operations console. Powered by a unified backend, shared data model, and observability layer, they automate critical freight workflows while keeping operations teams in complete control.
AI Rate Engine
Machine Learning (ML) model helps in analysing historical lane rates, surcharges, seasonal trends & live market signals to generate indicative freight quotes within seconds, associated with human reviews.
Document Intelligence
OCR & LLM-powered document parsing extracts structured shipment datasets from the Bills of Lading, invoices, and customs documents, along with a confidence-based validation workflow.
Predictive ETA
Here, machine learning combines carrier history, vessel tracking, weather & port congestion to provide continuously updated shipment arrival predictions with a confidence scoring system.
HS Code Classification
This involves a fine-tuned AI assistant that actually recommends ranked HS codes with duty implications, documentation requirements, & confidence scores while learning from operator confirmations.
Compliance & Regulatory Standards
Built around European freight regulations and customs requirements, with compliance modelled directly into shipment workflows, documentation, and operational data instead of relying on manual checks.
Engineering Stack We Shipped
Every technology selection was evaluated against three constraints: AI accuracy in a low-tolerance logistics environment, native bilingual experiences across customer touchpoints, and Rotterdam’s operational realities behind every shipment.
- Next.js
- React.js
- TypeScript
- Tailwind CSS
- Headless CMS
- i18n (next-intl)
- Locale-Aware Routing
- NL/EN Editorial
- XGBoost
- LightGBM
- Historical Lane Training
- Market Index Feeds
- Factor Attribution
- Azure Form Recognizer
- AWS Textract
- LLM Field Extraction
- Confidence Scoring
- Human-in-the-Loop Review
- ML Regression Model
- Port Congestion Signals
- Vessel Position Feeds
- Weather Integration
- Fine-Tuned LLM
- Customs Tariff Knowledge Base
- Embedding Search
- Confirmation Learning
- Node.js
- GraphQL
- REST APIs
- PostgreSQL
- pgvector
- Redis
- Carrier API Integrations
- Mapbox GL
- WebSocket Updates
- Milestone Engine
- AI Decision Logging
- Model Versioning
- Audit Replay
- Confidence Tracing
- AWS (Frankfurt & Amsterdam)
- NL Data Residency
- CloudFront CDN
- GitHub Actions CI/CD
How We Delivered the Build
We followed an AI-second development approach in the brand platform & operational data model, building first, as AI is reliable & useful as the freight data powering it. Stronger operational foundations ensured that every prediction was supported by reliable freight datasets.
Discovery, Workflow Audit & AI Priority Mapping
We mapped HJM’s commercial workflows, prioritized 4 AI capabilities, audited website gaps & also defined the bilingual content architecture.
Brand Site & Shipment Data Model
We built the bilingual brand site & unified shipment data model that covers EU customs, AEO, IMO, IATA, & BTW to give the later AI workload a trustworthy source.
AI Rate Engine & Quote Workflow
We trained the ML rate engine system associated with the platform utilizing historical lanes & market signals along with auditable quote workflows for operations.
Document AI & HS Classification Build
Established OCR, LLM document intelligence, HS classification, confidence scoring, & confirmation learning for custom processing workflows at enterprise scale.
Predictive ETA, Client Portal & Ops Console
We developed predictive ETA models, a client tracking portal & operations console system with total AI supervision & audit trails that support operations.
Observability, Bilingual QA & Launch
We incorporated AI observability, completed bilingual quality assurance, validated production models, & launched every platform component together.
What Our Client Says
Every engineering decision was driven by three imperatives: AI outputs had to remain explainable, freight workflows had to meet EU customs requirements, and bilingual operations had to perform consistently across every customer and operations touchpoint.
They didn’t just rebuild our website; they built the operational layer our ops team now spends their day in. Instant quotes, parsed documents, ETAs we can stand behind. AI applied to freight, not bolted onto a homepage.
We build applied AI for logistics: rate engines, document intelligence, predictive ETA, customs classification, alongside multimodal freight data models, EU customs compliance, and the bilingual depth Dutch operators need to win blue-chip procurement. Our experience as a logistics software development company enables us to build scalable, compliant, operations-first platforms for complex logistics ecosystems. Let’s talk.