SEMA
Semantic query engine for multimodal ETL, semantic ad-hoc querying, and unified analytics.
SEMA is a multimodal semantic data system designed to unify semantic understanding across structured and unstructured data. Powered by large language models, it targets the parts of data processing that traditional systems struggle to express cleanly.
It functions as a query engine for next-generation multimodal data processing, supporting semantic ETL for LLM pipelines, semantic ad-hoc querying over user requests, and analytics across text, images, video, and structured datasets.
The system goal is to let AI-heavy pipelines retrieve, interpret, and analyze heterogeneous content in one flow rather than stitching together separate systems for indexing, orchestration, and analytics.
Links: ArXiv paper