HIRI FUTURE’s port equipment automation vertical-domain large model has been selected as a high-quality industry-specific AI model in Fujian Province.

Recently, the Fujian Provincial Department of Industry and Information Technology officially released the “2026 List of Provincial Artificial Intelligence Projects.” Fujian Electronic Port Co., Ltd.—with its self-developed large-scale vertical model for port equipment automation under the brand “HIRI FUTURE”—was successfully included in the list of high-quality industry-specific AI models in Fujian Province. Only 10 companies in Xiamen made the cut. This selection signifies that the company’s years of dedicated R&D in port equipment automation AI technology, along with its practical industrial value, have received official provincial-level recognition, making it a benchmark achievement in Fujian’s port AI sector.

Model Overview and Applications

Unlike general-purpose large language models, the port‑equipment automation vertical‑domain large model is a specialized domain‑specific large model developed by HIRI FUTURE after years of focused research on smart ports and multimodal large‑model applications. Addressing key pain points in the port industry—such as low operational efficiency, difficulties in interoperability and collaboration among heterogeneous equipment, insufficient perception accuracy in complex scenarios, and high labor costs—the model is built around the core principles of “full‑domain coordination, scenario adaptation, and secure controllability.” It delivers intelligent perception, coordinated scheduling, and safety assurance across the entire port Procedure, making it the first vertical model system in China’s smart‑port sector to achieve an end‑to‑end intelligent closed loop encompassing “perception–decision–optimization–iteration.” By deeply integrating port‑wide business scenarios, this model overcomes longstanding industry bottlenecks—including data silos, heavy reliance on manual processes, low collaborative efficiency, and significant safety risks—thereby driving the transformation of port operations from “experience‑driven” to “data‑intelligent‑driven” and serving as the core intelligent foundation for fully unmanned operations throughout the smart port workflow.

At the underlying technical architecture level, the port equipment automation vertical-domain large model is built on the ViT (Vision Transformer) architecture as its backbone, establishing a full‑link technical pipeline that spans the “input tier–feature extraction tier–multi‑task output tier–closed-loop optimization tier.” The model employs a multimodal input encoding scheme to uniformly tokenize diverse data types, including images, text, and sensor readings. Through a self‑developed feature mapping module, it adapts the global features output by ViT into precise spatial features required for core tasks such as high‑precision positioning and object detection in port environments. Meanwhile, a proprietary multi‑task training framework has been constructed, supporting joint modeling with multimodal inputs and decoupled training across multiple tasks, enabling flexible adaptation to various task formats in port scenarios—such as perception, decision‑making, and autoregressive recognition—and ultimately forming a unified, efficient, and scalable multimodal modeling framework tailored specifically for ports.

Model training relies on the company’s self-developed high-quality dataset system for end-to-end automated port operations, which is one of the largest, most comprehensive in terms of scenario coverage, and highest in annotation accuracy among port-specific datasets in China, providing core data support for model training and iteration.

Market Applications

Currently, the model and its supporting solutions have been implemented at more than 80 major ports worldwide, covering over 3,500 systems in total. They serve China’s major hub ports, including Xiamen Port, Ningbo Port, Tianjin Port, Guangzhou Port, Shenzhen Port, Qingdao Port, and Shanghai Port, as well as ports in ASEAN countries such as Singapore, Thailand, and Malaysia, and in BRICS nations and ports along the Belt and Road, including Brazil and Sri Lanka. Several of these ports have already been integrated into the “Silk Road Maritime” Feeder route network, effectively enhancing port operational efficiency and safety, and strongly driving the transition of port operations from semi-automation to full-process unmanned operation.

Domestic substitution and industry demonstration have yielded significant results.

As the founder of the first smart port AI team among Fujian Province’s state-owned enterprises, “HIRI FUTURE” is dedicated to the in-depth R&D and industrial implementation of AI technologies in the port sector, achieving nationwide domestic substitution across key automation technologies—from core algorithms to computing power and front-end sensor hardware and software. This model has been deeply adapted to over 20 domestically produced chips and more than 50 domestic sensors, and its core solution has received authoritative certification from the China Academy of Information and Communications Technology, breaking the long-standing foreign‑technology monopoly in port automation and resolving critical bottlenecks in the industry. Leveraging these model‑based technological achievements, the company has successively been recognized as a top winner in the MIIT’s AI Industry Innovation Challenge (only two nationwide in the Intelligent Equipment category) and as a selected entity in the MIIT’s Future Industry Innovation Challenge (only two in Fujian Province in the AI field). Furthermore, the model was included in the first batch of Fujian Province’s key cultivation projects for vertical AI models in specific industries, establishing itself as a benchmark achievement in China’s smart port domain for large‑scale vertical models.