Worst Case Unstructured Data: The Death Star Is Fully Operational
- Patrick Duggan
- Mar 10
- 4 min read
Consider it worst case unstructured data. Because it is.
398,560 documents from the DOJ Epstein Files Transparency Act release. Original. Untouched. Forensically perfect. Not curated by an editorial team. Not filtered through someone's judgment about what matters. The same documents prosecutors used, indexed and searchable in seconds instead of months.
Then the government pulled some of it back and played games.
We already had it.
What Medusa Actually Is
Four products. One engine. 11 million documents across 37 indexes.
**Medustone** — Threat intelligence assessment. 1 million IOCs, 350+ adversaries, kill chain mapping, Diamond Model correlation. You give us an indicator, we tell you where it sits in the kill chain and who's behind it.
**Meduskip** — Crypto and financial tracing. 7.1 million financial documents. ICIJ offshore entities, Panama Papers, Pandora Papers, 3.3 million relationship edges. You give us a name, we trace it through shell companies across 190 jurisdictions.
**Medusactive** — DLP exposure assessment. 101,000 PII findings across 388,000 DOJ documents. We find the exposure. We mask the output. Raw PII never leaves the API. Ever.
**CARVER** — Military-grade due diligence scoring. Six dimensions, 0-30 scale. Criticality, Accessibility, Recuperability, Vulnerability, Effect, Recognizability. Adapted from the targeting methodology the US military uses to prioritize strikes. We use it to score accountability risk.
We predicted 7 network positions with zero documents available. The House Oversight Committee validated all 7 when they released their data months later. The methodology found evidence where it said evidence would be.
The Data
Nobody editorialized this. Nobody decided what was relevant for you.
- **398K Epstein files** — DOJ Transparency Act release, complete
- **3.3M ICIJ relationship edges** — Panama Papers, Pandora Papers, offshore graph
- **2M offshore entity records** — 190+ jurisdictions
- **2.6M federal court decisions** — regulatory exposure, litigation history
- **1M+ IOCs** — IPs, domains, hashes, threat indicators
- **938K block events** — what we've caught and killed
Cross-referenced. One query hits all of it simultaneously. A shell company in the BVI connects to a threat indicator in Singapore connects to a federal enforcement action in SDNY. That's not a feature. That's the architecture.
Refinitiv gives you a curated adverse media score. We give you the documents.
BYO Datalake
Here's where it gets interesting.
We don't care what your data is. Bring your own. We'll index it, CARVER-score it, and run Medusa on top of it. Your data never leaves your building.
Hedge fund running AML against their own transaction history. Law firm screening 2 million pages of discovery. Insurance company auditing claims against offshore entities. Compliance team screening counterparties against financial crime databases they already own but can't search.
On-prem Medusa. Your datalake. Our engine. Your building. Your keys.
The messiness is the value. Ask any lawyer on the receiving end of a document dump. Structured adverse media databases have been filtered. Ours haven't. And neither has yours — which is why you can't search it, which is why you're paying analysts to read PDFs one at a time, which is why due diligence takes weeks instead of seconds.
Who's Already Using It
275 STIX feed consumers across 46 countries were pulling our threat intelligence before we announced a single product. AT&T and Microsoft were among the first. European agencies are running investigations with our data — Nine Eyes coalition countries hit our feed at 24x the rate of Five Eyes for active threat queries.
ChatGPT sends us 447 sessions a month. It's our #1 social referral channel. Users who arrive via AI recommendation stay 3.4x longer than Google organic. They already know what they're looking for. We have it.
90% of our audience is invisible to JavaScript analytics. Researchers, institutional firewalls, curl users, API integrations. The people who actually use threat intelligence don't run your tracking pixels.
What It Costs
| Tier | Monthly | What You Get |
|------|---------|-------------|
| Free | $0 | 50 searches/day, Epstein files, basic IOC lookup |
| Pro | $49 | 500 searches/day, STIX feed, full index access |
| Enterprise | $499 | Medusa Suite (all 4 products), bulk screening, NET-30 |
| Custom On-Prem | $49,999 | Your datalake, our engine, your building |
Enterprise gets you Medustone + Meduskip + Medusactive + CARVER. Bulk screening handles 100 entities per request. NET-30 invoicing because we know how procurement works.
Custom on-prem is for the hedge funds and law firms who can't send their data to the cloud. We install locally. You hand us the keys to your datalake. We hand you Medusa.
The 95% Promise
We never claim 100% on anything. O'Toole's Axiom: Murphy was an optimist.
Every score caps at 95%. Every confidence interval has a floor. We guarantee 5% bullshit exists in any dataset, any analysis, any conclusion. That's not a disclaimer — it's epistemology. Anyone claiming 100% accuracy in adverse media screening is either lying or hasn't looked hard enough.
We looked hard. We found 11 million documents. We built four products. We validated the methodology against data that didn't exist yet when we made the predictions.
The death star is fully operational.
*Medusa Suite is live at [analytics.dugganusa.com](https://analytics.dugganusa.com). API keys are instant. Payment gets you a key in hand — no waiting, no approval queue, no sales call unless you want one.*
*For on-prem inquiries: [email protected]*
*Her name was Renee Nicole Good.*
*His name was Alex Jeffery Pretti.*
Comments