The demo will present a novel deep learning model for evaluating perceptual audio characteristics in the music domain intended for optimizing headphones and earbuds. The model is trained on extensive data collected by 35 expert listeners who evaluated generated signals for multiple audio characteristic attributes in the timbre domain, perceived loudness and personal preference on a great scale specifically designed for machine learning. The proposed model architecture and a detailed overview of the data and analysis methods will be presented. Attendees will be introduced to a web application designed to serve as an AI-driven assistant for tuning audio systems, providing visualizations and interpretable results from models trained via no-reference evaluations. The tool will be openly available and can help tonmeisters and others involved in the tuning of audio systems to more efficiently and consistently evaluate sound characteristics of their products. By providing a scalable, data-driven framework, this demo has the potential to set new benchmarks for evaluating sound quality and could inspire further innovation in AI-driven audio analysis. Attendees can interact with demo data from various samples and systems that mirror real-world listening tests, view spectrograms, and use the deep learning model within the web application to compare predictions with their own perception. If attendees have suitable datasets available, we encourage trying out the tool with their own data. We also welcome collaboration and feedback from the community, as the tool is intended to evolve based on user input. While the model is trained for the optimization of headphones and earbuds, we believe the proposed AI-assisted evaluation framework can easily be adapted to other domains, such as in the speech domain and the optimization of hearing aids.