DNS, WHOIS, and public-record aggregation with a PyTorch classifier labelling signals at 89% accuracy.
Open-source intelligence is scattered across DNS records, WHOIS registries, and public databases that share no format and no schema. A Scrapy spider network with custom middleware collects from all of them, handling the rate limiting and retries each source demands.
A PyTorch classifier then labels the collected signals, reaching 89% accuracy on a held-out test set. Everything lands in PostgreSQL with features normalised for model input.