Intelligent Virtual Assistant for Calculating Technology Readiness Levels Using Large Language Models (LLM)

Abstract

The Technology Readiness Level (TRL) is a metric developed by NASA in the 1970s to assess the maturity of a technology and guide decision-making regarding its development. Although it is standardized, in practice, evaluations are often manual and subjective, leading to inconsistencies. In this study, an intelligent virtual assistant was developed based on a large language model (LLM), using GPT-3.5-Turbo as the foundation, selected over LlaMa-2 after rigorous performance testing. The model was trained using optimization techniques on a custom dataset built from reliable sources such as NASA, government entities, and universities, enabling a more accurate evaluation of the TRL.