Voices from the Archive
A multimodal LLM platform that translates the 150,000-page multilingual Iraqi Jewish Archives, making them publicly accessible at judeoiraq.org. An 'Ask the Archive' Q&A interface lets readers ask plain-English questions and trace every answer back to the original page.
The Iraqi Jewish Archives are a 150,000-page collection recovered in 2003 from a flooded basement in Baghdad: religious texts, school records, business ledgers, government documents, photographs, letters, and ephemera spanning four centuries of Jewish life in Iraq. The pages were restored, scanned, and digitized. But they remained largely inaccessible to non-specialists: multilingual (Hebrew, Judeo-Arabic, Arabic, and English), often handwritten, and impossible to search at scale. Voices from the Archive opens the corpus to the public at judeoiraq.org, browsable, transcribed, translated, and searchable in English.
At the core is a multimodal LLM pipeline that ingests scanned page images, runs transcription and translation across multiple languages and scripts, indexes the results, and exposes them through both keyword search and an 'Ask the Archive' Q&A interface. The Q&A lets visitors ask plain-English questions and trace every answer back to the specific pages it drew from. Citation-grounded by design, so a reader can always check the source.
A solo, personal passion project. JIMENA is fiscal sponsor; the National Library of Israel stewards the digitized transcriptions and translations.
