Automated OSINT Platform
AletheiaHQ
Zero-Loss GraphRAG Pipeline
Built a fleet of distributed Python microservices that autonomously scrape, synthesize, and monetize highly fragmented government data, outputting step-by-step financial arbitrage directives.
[ architecture.snippet / aletheiahq ]
> (Chunk)-[:PART_OF]->(Document)
> vector_search(query) -> chunks
> traverse(graph) -> full_context
> rank: 1 (quick) .. 7 (monopoly)
[ GALLERY ]
[ TRANSLATION LAYER ]
// FOR_FOUNDERS
The Business Story
The Challenge
B2B clients need to find hidden financial opportunities buried in enormous piles of public government and market data — procurement filings, audits, grants — scattered across 8+ agencies in formats no human wants to read. By the time someone finds an opportunity manually, it's gone.
My Approach
I built an automated intelligence platform that scrapes, understands, and connects thousands of fragmented documents using a custom GraphRAG system, giving an AI a photographic memory of every filing plus the judgment to know which ones are worth acting on. When it finds something promising, it automatically researches the people and context behind it and writes a step-by-step plan for how to act on it.
The Result
The platform runs on autopilot, continuously ingesting new filings and producing ranked, ready-to-execute "Directives" — turning a task that used to take analysts weeks into something that happens automatically in the background.
// FOR_ENGINEERS
The Tech Specs
Stack
Engineering highlights
- >Designed a hybrid GraphRAG system that solves the classic "lost in the middle" LLM problem by linking localized vector chunks back to their parent document.
- >Built a resilient, Dockerized scraping fleet processing massive public datasets asynchronously without rate-limit blocking.
- >Bridged raw data engineering to business value: the pipeline doesn't just parse data, it writes the commercialization strategy for it.