AI Logistics Cloud

Domain AI Capability, Built From Real Freight Operations

AI Logistics Cloud is our industry AI platform, built on the scenarios we handle every day: forwarding, customs documentation, yard and warehouse operations, and electronics distribution. Customers access domain-specific models, GPU cloud compute, fine-tuning and training services, hosted AI agents, token governance and multimodal token metering through one platform.

AI Logistics Cloud GPU Cloud Service Multimodal Token Service

Platform Positioning

From Internal Upgrade to Industry Platform

Our AI work started inside our own operation, where the problems are concrete: incomplete enquiries, re-keyed documents, disputed cargo condition, and safety supervision that depends on who happens to be walking the yard. Solving those problems produced models, agents and governance rules that other companies in the same industry face as well.

The platform makes that capability available to industry customers. It is built for logistics, forwarding, trading and warehousing scenarios, with domain terminology, document formats and operating rules already reflected in the models, and with the ability to adapt further to a specific customer's data.

Domain models for logistics scenarios

GPU cloud for training and inference

Fine-tuning on customer data

Hosted agents with audit trail

Token metering and access control

Multimodal token metering and billing

Five-Layer Architecture

Layer 5

Service Layer

Open API, customer console, token metering and billing, private or hybrid deployment options for customers with data residency requirements.

Layer 4

Capability Layer

Agent orchestration, visual recognition, document parsing, forecasting and scheduling, exposed as reusable building blocks.

Layer 3

Compute Layer (GPU Cloud)

Training, fine-tuning and inference capacity scheduled by job priority, with elastic scaling for online services and private compute options where data cannot leave customer premises.

Layer 2

Model Layer

General foundation models combined with industry-specific models trained and fine-tuned on logistics, customs and warehouse scenarios.

Layer 1

Data & Knowledge Layer

Industry corpora, document samples, handling imagery and operational records, governed with access control and retention rules.

Model Hub

Industry AI Models Available to Customers

Each model is delivered through API or private deployment, with maturity stated honestly. Evaluation results are reported per scenario against an agreed evaluation set.

Logistics Document Recognition

Pilot

Commercial invoice, packing list, bill of lading and customs declaration parsing

Input: scanned or photographed documents. Output: structured fields with confidence scores.

API / Private deployment

HS Code & Declaration Assistant

Pilot

Classification review and declaration element checking before submission

Input: product description and specifications. Output: candidate HS codes with reasoning and required elements.

API

Freight Rate & Capacity Forecast

Planned

Lane-level rate trend and equipment availability outlook for booking decisions

Input: lane, period, cargo profile. Output: indicative range and confidence level.

API / Console

Cargo Visual Condition Recognition

Pilot

Container and pallet condition, seal verification, stacking and lashing anomalies

Input: yard or handling point imagery. Output: condition labels, anomaly flags and time-stamped records.

Private deployment

Multilingual Customer Service Model

Live

Enquiry handling and status replies in English, Bahasa Melayu and Chinese

Input: customer message and shipment context. Output: drafted reply for officer review.

API / Agent Cloud

Warehouse Safety Behaviour Recognition

Planned

PPE compliance, forklift and pedestrian conflicts, restricted zone intrusion

Input: fixed camera streams. Output: event records with imagery for officer confirmation.

Private deployment

Platform Capabilities

Fine-Tuning, Agent Cloud and Token Cloud

Three services that turn industry models into something a customer can actually run: models adapted to their data, agents that execute defined tasks, and a metering layer that keeps usage and cost under control. GPU cloud compute and multimodal token metering are documented on their own pages.

Fine-Tuning Service

Industry Model Fine-Tuning & Training

For customers whose terminology, document formats or operating rules differ from generic models, we provide an end-to-end fine-tuning service: from data assessment to evaluation, deployment and ongoing iteration.

Requirement diagnosisData preparationTraining & evaluationPilot validationDeployment & operations

Standard Delivery Flow

Requirement diagnosis→
Data preparation→
Training & evaluation→
Pilot validation→
Deployment & operations

Data Assessment & Cleaning

Review sample volume, quality and coverage, then define what additional data is worth collecting before any training starts.

Annotation Standards

Written labelling rules for documents, cargo conditions or safety events, so results stay consistent across annotators and batches.

SFT / LoRA Fine-Tuning

Supervised fine-tuning and parameter-efficient adaptation on industry corpora, with version control for every training run.

Evaluation Set & Acceptance

A held-out evaluation set agreed with the customer, with metrics reported per scenario rather than a single averaged score.

Knowledge Injection (RAG)

Customer SOPs, tariff rules and service scopes connected as a retrieval knowledge base so answers stay grounded in approved material.

Continuous Iteration

Production feedback and corrected cases flow back into the next training cycle, with rollback available at version level.

Agent Cloud

AI-Agent Service Cloud

A hosted environment where industry customers compose, run and govern task-oriented agents. Agents call your systems through defined interfaces, and every action is logged with a human takeover path.

Agent template libraryTool & API integrationMulti-agent orchestrationPermissions & audit trailHuman-in-the-loop takeoverCanary release

Quotation & Enquiry Agent

Captures cargo details, checks completeness, retrieves tariffs and prepares a structured quotation draft for the desk.

Document Processing Agent

Parses incoming documents, maps fields into your workflow and flags inconsistencies before submission.

Shipment Tracking Agent

Consolidates milestone events and answers status questions in the customer language, escalating exceptions to an officer.

Customer Service Desk Agent

Handles first-line enquiries across channels with knowledge base grounding and defined escalation rules.

Safety Inspection Agent

Reviews visual detection events, groups repeated findings and prepares a corrective action record for review.

Supply Chain Analysis Agent

Turns operational data into periodic reviews of transit performance, cost structure and exception patterns.

Token Cloud

AI-Token Cloud & Usage Governance

A single gateway and metering layer across models, so industry customers can control spend, attribute usage to the right project and keep an auditable record of every call.

Per-project meteringBudget alertsBilling exportModel version pinningRate limiting

Unified Model Gateway

One API surface across multiple models, with routing rules per scenario and fallback when a provider is unavailable.

Quota & Allocation

Token budgets assigned by project, department or end customer, with hard limits and soft warnings.

Cost Visibility & Alerts

Usage and cost dashboards with threshold alerts, so overruns are noticed during the month rather than at billing.

Caching & Routing Optimisation

Repeated prompts and stable reference material served from cache, and lighter models routed to simpler tasks to reduce cost.

Logging & Audit

Call-level records with request context, model version and outcome, retained for a defined period and exportable for reconciliation.

SLA & Rate Limiting

Throttling and priority queues to protect production workloads during peak usage.

Dedicated Service Pages

GPU Cloud and Multimodal Token Services

Two platform services are documented on their own pages, with capacity tiers, counting rules and governance controls described in enough detail to plan a workflow against them.

GPU Cloud Service

Elastic inference nodes, fine-tuning queues, multi-accelerator training capacity and batch throughput pools, with priority scheduling, checkpointing, utilisation reporting and private compute deployment.

View GPU Cloud Service

Multimodal Token Service

Text, image, video and voice usage measured on their own terms, converted into one comparable unit, with cross-modal task tracing, per-modality quotas and modality-level audit logs.

View Multimodal Token Service

Industry Applications

Where the Platform Fits in Real Operations

Each scenario below pairs a common operational difficulty with the model and agent combination we would propose, and the direction of benefit we would measure during a pilot.

Freight Forwarders

Typical Difficulty

Enquiry volume spread across email, phone and messaging, with quotation turnaround limiting how many lanes can be served.

Proposed Combination

Document Recognition + Quotation Agent + Token Cloud

Direction of Benefit

Faster first response and consistent quotation structure across branches.

Port & Yard Operators

Typical Difficulty

Safety supervision depends on patrol coverage, and condition disputes are hard to evidence after the fact.

Proposed Combination

Cargo Visual Condition + Safety Behaviour Recognition + Safety Inspection Agent

Direction of Benefit

Documented condition records and earlier detection of unsafe practice.

Cross-Border E-Commerce & Traders

Typical Difficulty

High SKU counts with frequent catalogue and compliance changes across destination markets.

Proposed Combination

HS Code Assistant + Multilingual Service Model + Tracking Agent

Direction of Benefit

Fewer classification corrections and clearer customer status answers.

Manufacturing & Project Logistics

Typical Difficulty

Oversized and staged deliveries involve many parties, so handover responsibility is often unclear.

Proposed Combination

Document Recognition + Tracking Agent + Supply Chain Analysis Agent

Direction of Benefit

One custody record per leg and periodic review of exception patterns.

Third-Party Warehousing

Typical Difficulty

Inbound and outbound checks are manual, and stock accuracy depends on cycle counting discipline.

Proposed Combination

Cargo Visual Condition + Safety Behaviour Recognition + Token Cloud

Direction of Benefit

Visual stocktaking support and monitored handling practice.

Not sure which combination applies?

Send us a sample of your documents, a description of the workflow or a site layout. We will review it and propose a bounded pilot with success criteria agreed in advance.

Discuss a pilot

Security & Compliance

Controls We Apply Before Scale

Data Isolation & Least Privilege

Customer data is separated per tenant, and access is granted only to the roles that need it for a defined task.

Retention & Access Windows

Imagery, documents and call logs are kept for an agreed period, with access limited to authorised staff and signage displayed at monitored areas.

Regulatory Alignment

Handling of personal data follows Malaysian personal data protection requirements, and cross-border transfer arrangements are confirmed before deployment.

Human Review of Model Output

Model results support decisions rather than replace them. Declarations, quotations and safety findings are confirmed by a responsible officer.

Honest Performance Reporting

Evaluation results are reported per scenario with the evaluation set disclosed. We do not claim absolute accuracy for any model.

Deployment Flexibility

Where data cannot leave customer premises, models and agents can run in a private or hybrid arrangement.

Platform Roadmap

Built in Stages, Opened in Stages

Near Term

Model API, GPU Inference & Token Metering

Document recognition and multilingual service models available through API
GPU inference capacity with elastic scaling for customer endpoints
Token gateway with per-project metering and budget alerts
Sandbox environment for customer evaluation

Mid Term

Agent Cloud, Multimodal Token & Fine-Tuning

Hosted agent templates for quotation, documents, tracking and safety review
Multimodal token metering across text, image, video and voice workloads
GPU training and fine-tuning queues for industry engagements
Private and hybrid deployment options

Longer Term

AI Logistics Cloud Ecosystem

Shared industry knowledge bases contributed by participating customers
Cross-scenario agent orchestration across forwarding, agency and trading
Partner integration for carriers, warehouses and customs service providers

Capability Trial

Sandbox access to selected models so your team can test against real samples.

Scenario Pilot

One bounded workflow, with success criteria agreed before the pilot starts.

Joint Fine-Tuning

Your data and terminology incorporated, with evaluation results reported per scenario.

Platform Subscription

Ongoing access to models, agents and token governance under an agreed service level.

Request a Platform Capability Session

Tell us the workflow you want to improve and the data you can provide. We will propose a bounded pilot, the models and agents involved, and the evaluation criteria we would report against.

Contact Our Team