• Progress: Set up the uv environment to test the PDF table extraction using pdfplumber (after trying pypdf and xberg). Used SWE 1.7 inside Devin Desktop IDE to build an end-to-end Python pipeline that extracts, normalizes, and loads table data into a DuckDB database.

  • Struggle: Extracted data quality is inconsistent. The San Diego reports have complex, handcrafted layouts with merged columns and multi-layered headers. Comparing two different report years revealed some format drift, the newer report contains more data and sometimes a completely different layout, making reliable parsing difficult and hard to trust, but not impossible to deal with.

    • 2026

    • 2007

  • Plan: Put together a 4-5 minute demo video and status update to showcase what was accomplished this week, outline the extraction challenges, and present potential next steps for the project.