Customers & Partners

Background

Massive data accumulation, scattered expertise, complex processes, diverse tools and systems, volatile markets and faster product iteration — all routine for this manufacturing giant. The company set out to build a systematic AI capability and one unified AI enablement platform.

Challenges

• Intelligent interaction and tool integration:Build a unified human–machine entry point that supports multi-turn natural language dialogue and language switching.
• Knowledge services and KBE integration:Aggregate heterogeneous knowledge from multiple sources into an ontology-based knowledge hub.
• Data management and intelligent analysis:Consolidate on-premise and online BOM data, with fully private deployment and permission control.

Solutions

• Intelligent conversation layer:A unified interaction entry point that fuses knowledge retrieval with business execution, supporting cross-language interaction compliant with KBE coding standards.
• Cross-platform collaboration layer:Connects business applications horizontally across domains, integrates infrastructure and data resources vertically, and establishes cross-functional collaboration.
• Data management layer:With BOM information as the data thread, AI empowers every link — R&D, manufacturing, sales and service.

Results

Intelligent scenarios and applications including AI4CAD and AI4CAE have gone live, establishing a paradigm that fuses AI technology with deep domain work and delivering core capabilities such as a knowledge hub driven by intelligent technology, KBE integration and BOM comparison. InLower design cost, higher production efficiency, better verification quality, new business valueAll four dimensions delivered significant results, giving the industry a lighthouse case and a reference approach for AI-empowered manufacturing.

Background

During its digital transformation, the company faced severe information silos: design, manufacturing and quality inspection each ran on separate systems with inconsistent formats and rigid barriers. Cross-department work required many conversations and many system switches — slow, laborious, and decisions were made on incomplete evidence.

Challenges

• Data barriers make collaboration hard:R&D parameters, digital models and drawings rarely reach process engineering, so data consistency across domains cannot be guaranteed.
• Process knowledge is hard to reuse:Blade process design relies on individual experience with no systematic capture — hard to search, hard to reuse.
• Advanced technology is hard to absorb:Inefficient process practices block new technologies and new processes, and hold back the growth of process engineers.

Solutions

• Intelligent application layer:Built on large models, agents and MCP to enable intelligent generation, recommendation and prediction — powering blade information retrieval, process route generation, UID generation and out-of-tolerance risk detection.
• Knowledge base layer:Using RAG and knowledge graphs, build an AI knowledge base where a graph database and a vector store work together, achieving full-scope semantic fusion and unified storage.
• Cross-system fusion layer:Design large-model-based cross-system interface protocols and presentation components to integrate data and functionality across heterogeneous systems.

Results

Build a visual association network spanning the full lifecycle of turbine blades, connect seamlessly with multiple heterogeneous systems, and remove business data barriers entirely — one search reaches through the whole chain across departments. The platform supports real-time multi-user retrieval, and an innovative timeline surfaces key events along the blade lifecycle, giving strong support to intelligent scheduling.

Background

Quotation in pharma CDMO spans requirement parsing, route selection, cost accounting and commercial strategy, with information scattered across departments and historical documents. It leans on veteran experience, responds slowly and lacks a consistent basis.

Solutions

• Requirement parsing:Structure the customer's RFQ documents into computable requirement elements.
• Route recommendation:Recommend feasible synthesis routes from historical projects and the process knowledge base.
• Cost accounting:Link material, labour-hour and capacity data to generate a cost range quickly.
• Strategy optimization:Give quotation recommendations against margin targets and the competitive landscape.