[Carver Under The Hood] Public Context Infrastructure - that’s what we’ve been building.
Jaya Gupta’s recent piece and Aaron Levie’s comments gave us better language for something we’ve been working on over the past year. We’ve been building context graphs specifically for public data, and what we learned is that it’s not just about the graph - it’s about a full stack for the public data half of every company’s context graph.We expose this as APIs and services that plug directly into existing data platforms, applications, and agents. While we started with regulatory data, the same infrastructure naturally extends to other public data domains over time.
We expose this as APIs and services that plug directly into existing data platforms, applications, and agents. While we started with regulatory data, the same infrastructure naturally extends to other public data domains over time.
We had three key learnings that took us beyond the basic graph:
1. Domain Specificity
We found that you can’t build these graphs in a generic way. Each context graph needs to embed deep domain knowledge - legal interpretations, regulatory workflows, corner cases, and how obligations actually get interpreted inside teams. In our case, that means weaving a lot of public legal and regulatory understanding directly into the graph.
2. Continuous Learning
These graphs aren’t static. They need a set of agents and services to continuously maintain and enrich them - filling gaps, scoring trust, adding third-party context, and keeping sources reliable. This becomes an intelligence layer that keeps the graph evolving continuously and reliably.
3. Decision and Modeling Infrastructure
Finally, the graph alone isn’t enough. We need a decision-making and modeling layer on top of it - impact analysis, prioritization, scenario assessment, and downstream actions - because the ultimate use of the graph shapes how it should be built in the first place.
If you’re working with public data, or thinking about incorporating public data into your context graphs, let’s talk.