This post discusses the integration of Rust for PDF parsing within Elixir applications, particularly for RAG (Retrieve and Generate) capabilities essential for leveraging LLM (Large Language Models). While simple text extraction can be achieved using existing utilities like `pdftotext`, the real challenge lies in accurately extracting complex formats such as tables. There is a need for high-performance tools that can handle these tasks effectively, drawing parallels to existing tools like Unstructured and Marker. However, the use of native code (NIFs) in the BEAM architecture raises concerns about reliability, prompting the idea of isolating native processes into separate VMs to enhance stability. Additionally, there is interest in hybrid systems that combine LLMs with visual processing capabilities to enrich data extraction capabilities.