Energy Efficiency Project: Knowledge Retrieval System
In partnership with Thinkfox.ai
The problem: Finding state-specific info quickly is daunting for energy efficiency consultants
Energy Efficiency consultants often face the daunting task of navigating hundreds of pages of technical regulations from multiple states. These documents, known as Technical Reference Manuals (TRMs), are complex, spanning over 1000 pages with detailed charts, equations, and diagrams.
TRMs vary significantly by state and cover both commercial and residential building codes, making it challenging to locate specific information quickly.
The solution: An advanced RAG (retrieval-augmented generation) system to efficiently find relevant info
Our solution was to develop an AI-powered Retrieval Augmented Generation (RAG) system that transforms this cumbersome process into a streamlined experience.
By enabling consultants to ask natural language questions like, “What factors are needed to calculate energy efficiency for a warehouse in Albany?”, the system rapidly pulls relevant content from these vast resources.
The core of our work lay in optimizing data preparation for the RAG system.
We focused on converting PDFs, chunking data for effective retrieval, generating embeddings, enriching metadata, and fine-tuning a vector database for peak efficiency. Advanced technologies like stepback query, query rewriting, and query routing were employed to maximize the system’s capability to deliver the most relevant and precise answers.
This project showcases how generative AI can simplify regulatory complexity and empower energy consultants to make informed decisions efficiently.