AAI_2025_Capstone_Chronicles_Combined

‭ResolveAI‬

‭Desired State‬

‭The final product will incorporate multiple AI-driven components designed to optimize‬ ‭ticket resolution workflows. At its core, a Retrieval-Augmented Generation (RAG) component‬ ‭will retrieve relevant past tickets or documentation from a vector store or database (e.g., Chroma,‬ ‭Elastic, FAISS), ensuring that responses are grounded in existing knowledge and previous‬ ‭resolutions. We plan on adding a classification component that will leverage a bidirectional‬ ‭LSTM architecture to classify the priority of incoming requests/tickets based on the type of ticket‬ ‭and the content of the ticket (Team, K., n.d). Finally, the system will generate responses by‬ ‭integrating LLM capabilities with RAG, ensuring that answers are not only contextually accurate‬ ‭but also tailored to the specific needs of each ticket. This multi-layered approach enhances‬ ‭efficiency, minimizes manual intervention, and ultimately accelerates resolution times while‬ ‭maintaining high-quality support.‬

‭Dataset Summary‬

‭Dataset Introduction‬

‭Our dataset contains several key columns related to helpdesk or ticketing scenarios.‬ ‭Notable variables include:‬

‭●‬ ‭subject (text):‬ ‭Contains the ticket's subject line. Missing values are imputed with “[No‬ ‭Subject]” to retain data integrity.‬ ‭●‬ ‭body (text):‬ ‭Provides the detailed description of the issue. Records with missing body‬ ‭values were removed.‬

‭5‬

53