MissingIntel turns decades of fragmented public investigation records — police dossiers, court documents, journalistic archives — into a searchable, interconnected knowledge graph. Starting with one of the most documented cases of our time.
Built on the same retrieval and reasoning techniques that power frontier AI systems — adapted for the messy reality of historical case files: scanned PDFs, multilingual transcripts, decades-old witness statements.
Ask questions in natural language. Retrieval-augmented generation surfaces relevant passages from across the entire corpus, with citations back to the original page in the original source.
Persons, locations, vehicles, and timelines are extracted and linked. See at a glance who met whom, who was where on which night, and which threads have been buried in the noise.
The system surfaces statistical similarities between persons, places, and patterns — and flags them as hypotheses, never conclusions. The reader judges. The platform merely makes the connections findable.
Few missing-persons cases have generated as much public documentation. The complete Portuguese Polícia Judiciária file — 17 volumes, tens of thousands of pages, made public in 2008 — forms the spine of our pilot dossier, supplemented by court records, media archives, and witness timelines from the past 19 years.
Every page of the original Portuguese investigation has been OCR'd, translated to English and Dutch, indexed semantically, and structurally linked into a knowledge graph.
Search any name, any night, any vehicle license plate. Read the original Portuguese alongside the translation. Follow a witness through every appearance across the corpus.
Every answer the platform produces is grounded in retrieved source documents and accompanied by citations. No black-box claims. No invented facts. If we can't find it in a public record, we don't claim it.
Scanned pages are processed with high-accuracy OCR, translated where needed, and embedded into a multilingual vector store. Persons, locations, dates, and vehicles are extracted into a graph.
Queries trigger both semantic similarity search and structured graph traversal. Top passages are ranked, deduplicated, and assembled into context — always with traceability back to source.
A frontier reasoning model synthesizes the retrieved evidence into an answer. Citations are inline. Speculation is marked. Statistical similarities are presented as hypotheses, never as proof.
The Madeleine McCann dossier is the first public case on MissingIntel. The platform is designed to host additional cold cases, historical disappearances, and unresolved investigations over time — under the same standards of transparency, sourcing, and respect for due process.
Launch Platform →