CECI Proposes Engineering Skills Framework to Turn R&D Assets into Actionable Agent Modules
AI assistants · Enterprise AI · Structures
CECI Engineering Consultants explored a paradigm shift in knowledge management by encapsulating engineering reports into actionable 'Engineering Skills'. Leveraging RAG and Agentic AI, static documents are converted into reusable, executable capabilities, demonstrated through bridge seismic assessment workflows.
According to a feature report in CECI Engineering Technology by CECI's BIM Integration Center, engineering consulting firms are shifting knowledge management from static document archiving to AI-driven capability deployment. Based on 23 internal R&D projects completed in 2025 across infrastructure resilience, automation, sustainability, and advanced technologies, the study addresses the limitations of manual file retrieval. CECI designed a three-layer AI matching and retrieval framework covering knowledge assets, translation/governance, and applications. Utilizing Retrieval-Augmented Generation (RAG) and Agentic AI principles, technical reports and domain expertise are transformed into standardized, executable 'Engineering Skills' comprising metadata and code. Demonstrated in a bridge seismic preliminary evaluation use case, calling parameterized assessment skills enabled rapid iterative comparisons, while visual translation skills rendered abstract damage indicators into interactive 3D failure models to assist decision-making.
Key takeaways
- CECI utilized 23 internal R&D projects from 2025 to encapsulate domain knowledge into modular 'Engineering Skills' via RAG and agent architectures.
- A three-layer AI matching and retrieval platform was constructed, establishing a standardized pipeline featuring quality auditing and an Agent Factory.
- A bridge seismic assessment toolkit demonstrated that parameterized agile iteration and 3D damage visualization accelerate design comparison and decision-making.
Source：中華顧問工程司／台灣世曦｜中華技術 Read the original ↗
Prepared, checked and automatically published by AI; not individually reviewed by a human. Translation checked by AI.