Use of AI technologies for Specification Review and Cost Analysis of Transportation Projects

Abstract: Transportation construction projects vary in size from a few thousand dollars as managed by a local township to a mega project of billions of dollars as managed by a State Department of Transportation (DOT) or Federal Agency. In between are construction projects of every size, specialty, and contracting method imaginable. Success of a transportation project requires close collaboration of engineering firms, contractors, and owners using data that is reflective of the conditions at the project location during construction. AI based technologies can bring increased transparency and collaboration among these stakeholders by leveraging comprehensive data sets, including historical cost information. This paper presents innovative AI-based methodologies that leverage Large Language Model (LLM) capabilities to deliver unified, real-time technical specifications and historical cost information for transportation project stakeholders.

This paper will present AI based tools that will help users effectively identify special provisions, technical specifications, manuals, and guidelines (contract documents) that control a given project. The contract documents of a given agency go through updates, and controlling contract documents will vary considerably due to factors such as date of letting, contracting method, controlling agency, and project specific requirements. This paper will present AI based tools that will help contractors, designers, and owners navigate the contract documents to identify the requirements for a specific product or construction method that needs to be incorporated into a given project. AI based methodologies will help users evaluate the unit costs for the product for a similar application for the same agency. This paper will present intelligent methods that can be used to link the projected unit cost for a product or a construction method with the historical cost data that is available based on the location of the project and quantity required.

This paper will demonstrate how to leverage fine-tuned LLM with deep domain specific understanding integrated Retrieval Augmented Generation (RAG) for precise semantic information retrieval for user queries. The paper will explore how a fine-tuned large language model enhanced with Retrieval-Augmented Generation improves the performance of customized foundational models by incorporating dynamic external knowledge repositories into their responses. Additionally, it will discuss methods for comparing real-time costs of the projects with historical cost data. The LLM architecture proposed in this paper will allow access to a broad domain specific knowledge base, enabling stakeholders to tap into ever changing, vast, and diverse datasets, including project-specific documentation, industry standards, and best practices, and will provide more precise and contextually informed responses to queries, ensuring that team members have access to the most accurate and up-to-date information. In short, this paper will present how cutting-edge AI solutions will enable users to quickly access the most relevant, up-to-date information tailored to their specific project requirements.
This paper will also summarize case histories where an AI-driven approach improved project outcomes, enhanced stakeholder collaboration, and increased productivity. This paper will conclude by demonstrating that AI based tools automate much of the information gathering and analysis process, saving time and reducing the mental effort required from team members.