Integrating and Enhancing Building Data for Advanced Research: NEST-Bot

The built environment generates complex and heterogeneous data, categorized into 3 main types: structural and architectural information, performance data (time series of energy consumption, temperatures, or occupancy), and administrative records (contracts, costs). Despite the critical value of ORD in fostering scalable applications, significant challenges persist, including fragmented data storage, heterogeneity in standards, and inadequate metadata documentation, which complicates data contextualization and accessibility. This project, NEST-Bot, aims to address these challenges by enhancing data discoverability through an automated integration layer that populates a knowledge graph. NEST-Bot will train a LLM to serve as an intuitive interface for stakeholders-ranging from academic researchers and data scientists to HVAC engineers, architects, and automation experts-to access NEST-related data. A key aspect of the project involves the automatic generation of the integration layer from existing repositories, allowing the LLM to retrieve complex, heterogeneous datasets via natural language queries. This innovative approach aims to streamline data retrieval, enhance data quality, reduce redundancies, and make ORD practices more scalable and beneficial to the building sector. By linking and organizing diverse repositories, NEST-Bot will enable seamless interaction with complex datasets, establishing new standards for data integration and ORD in building automation and research.