While hundreds of performances take place every month in the classical music market, information is fragmented across platforms such as KOPIS, individual venue websites, and personal blogs. For novice audiences or those just beginning to broaden their tastes, finding a "performance they might like" presents a significant barrier to entry. Classical music fans manage their viewing history sporadically through notebooks, memos, and social media posts. Ticket photos get buried in galleries, and memories of which pieces were heard or which performers were present fade over time. Since this viewing data is not systematically accumulated, personalized recommendations based on that data have been impossible. This service allows classical music audience members to simply take a photo of their ticket, and AI automatically recognizes and archives performance information. Based on accumulated viewing history and taste profiles, it then recommends hyper-personalized performances through a three-stage AI pipeline (Multi-Query Retrieval → Reranking → LLM-as-Judge).

While hundreds of performances take place every month in the classical music market, information is fragmented across platforms such as KOPIS, individual venue websites, and personal blogs. For novice audiences or those just beginning to broaden their tastes, finding a "performance they might like" presents a significant barrier to entry. Classical music fans manage their viewing history sporadically through notebooks, memos, and social media posts. Ticket photos get buried in galleries, and memories of which pieces were heard or which performers were present fade over time. Since this viewing data is not systematically accumulated, personalized recommendations based on that data have been impossible. This service allows classical music audience members to simply take a photo of their ticket, and AI automatically recognizes and archives performance information. Based on accumulated viewing history and taste profiles, it then recommends hyper-personalized performances through a three-stage AI pipeline (Multi-Query Retrieval → Reranking → LLM-as-Judge).