In Q2 2023, a small update buried in a Taiwan-listed supplier’s investor section noted:
“We have expanded component supply for high-frequency satellite payloads under a North American aerospace contract.”
Most vendors ignored it.
RedGraphs didn’t.
We:
Disambiguated the supplier to a niche Taiwanese RF module company,
Mapped its direct customer to a U.S. Tier-1 defense integrator,
Inferred the end customer — Lockheed Martin, via multi-hop reasoning and past disclosures,
Estimated flow-through revenue based on component type, contract history, and supplier segment growth,
Anchored the relationship in time,
And flagged the link as part of a geopolitical decoupling signal — a sourcing shift likely driven by U.S.–China risk policy.
This surfaced months before it appeared in procurement trackers, analyst notes, or legacy data platforms.
It’s not just a link. It’s a timestamped, directional, financially weighted commercial relationship, with geopolitical context and source-level attribution.
The real structure of the economy isn’t described by sectors.
It’s described by who buys from whom, for what, how much, and for how long.
But that structure lives in:
Footnotes and filing attachments,
Regulatory disclosures and ESG statements,
Multi-lingual press releases and fragmented affiliate networks.
We’ve built the only system in the world that extracts, reasons over, and models those relationships in near real time.
RedGraphs combines LLM reasoning, transformer extraction, and proprietary temporal modeling into a scalable relationship intelligence engine.
We train domain-specific transformer models to:
Extract commercial entities, roles, contracts, and financial indicators
Parse sector-specific phrasing (e.g., “volume guarantee”, “design win”, “offtake agreement”)
Distinguish subtle directional signals that define real money flows
RedGraphs fine-tunes large language models on millions of labeled financial and operational documents. They:
Resolve aliasing across affiliates and geographies
Connect fragmented disclosures over time and across languages
Infer unspoken economic roles (e.g., indirect customers)
Generate natural language justifications for every relationship
These aren’t simple prompts. Our LLMs operate as multi-step agents, chaining evidence across time and documents.
Most filings don’t tell you when a relationship began or ended. We model it anyway.
Our temporal inference engine:
Aligns relationships to fiscal timelines and contract cycles
Resolves conflicting or missing dates
Detects quiet continuations, latent decays, and renewals
Anchors links to real-world activity periods, not just mentions
This allows RedGraphs to model the evolving lifecycle of each relationship — not just whether it existed.
Each relationship in the RedGraphs network includes:
Role direction (supplier → customer)
Timestamps (start, active, end)
Confidence score and explanation
Source quote span with link
Estimated economic exposure (if available)
Everything is explainable, traceable, and machine-usable — ready for risk models, alpha pipelines, or policy diagnostics.
Our graph is already enabling:
Investment teams to model supply and customer concentration risk across 700,000+ companies
Geopolitical analysts to track defense and critical tech sourcing shifts
ESG and compliance leads to trace sourcing exposure across opaque multi-tier structures
Data buyers to get ahead of disclosures, not wait for them
If you care about:
Defense supply chain security
Semiconductors and nearshoring
EV battery sourcing
Pharma contract manufacturing
Rare earths and global decoupling
Or simply want early signals of commercial change…
Then you’re either using RedGraphs or competing against someone who is.
We’ve mapped global money flows. Everyone else is still guessing.