bidlyzer.ai
Bidlyzer catches fraud in US government contracts.
Core trick
Runs RAG over FAR/DFARS clauses and fuzzy-matches contractor and officer names against ICIJ offshore entities, with LLM validation of matches.
Description
Bidlyzer catches fraud in US government contracts. The government hands out $765B+ in contracts every year and fraud is estimated around $500B but nobody investigates because its too expensive. One contract review takes a legal team weeks.
We built a 5-agent pipeline with LangGraph and Claude that takes a contract PDF and flags it in under 2 minutes. Agent 1 parses the PDF. Agent 2
runs RAG against federal acquisition regulations (FAR/DFARS) to find violations. Agent 2.5 compares the contract against peers from USASpending.gov so you know if a red flag actually matters in context. Agent 3 checks every contractor and officer against the ICIJ Offshore Leaks
database (Panama Papers, Paradise Papers, 814K entities) with fuzzy matching and LLM validation to cut false positives. Agent 4 generates an
investigation dossier with False Claims Act theories and evidence chains with page numbers.
90.6% accuracy on a 32-doc benchmark of real Federal Register procurement documents. The demo video shows an earlier prototype, we've since added
the baseline comparator agent and ICIJ entity resolution based on 3 months of research we compiled during this build.
The linked repo returns 404, so it was deleted or made private.