Skip to content
Research

Graph-first infrastructure intelligence (research)

We build research prototypes for mapping dependencies and exploring cascading risk across complex systems. The emphasis is decision support and analysis.

Directed graphs Dependency mapping Scenario planning Cascading risk

Dependency graph modeling

Build graph representations of systems: entities as nodes, relationships as edges, and queryable paths that explain “what depends on what”. This work is research-first and intended for analysis and decision support.

  • Ontology + schema design (kinds, relationships, provenance)
  • Graph queries for exposure / dependency / choke-point reasoning
  • Uncertainty + provenance labeling (what is known vs inferred)
  • Reproducible experiments and evaluation
  • Prototype UI for exploration

Lens + retrieval prototypes

Prototype “lenses” that slice a graph/map into a focused view, and retrieval components that attach relevant context and sources.

  • Lens APIs (scoped views with budgets, filters, and provenance)
  • Source attribution + evidence panels
  • Retrieval experiments for reports / signals (research-first)

Data ingestion + normalization

Build ingestion pipelines that fetch, normalize, cache, and serve third-party datasets for analysis.

  • Dataset adapters with caching + TTL
  • Normalization + dedupe + ID strategy
  • Offline bundles for reproducible research

APIs for research tooling

Simple APIs that serve datasets and lens outputs to interactive demos and notebooks.

  • REST/JSON endpoints for lens execution
  • Dataset status + refresh controls
  • Reproducible exports (GeoJSON/JSON)

Interactive demos

Human-in-the-loop interfaces for exploring networks, testing scenarios, and inspecting provenance.

  • Map + graph exploration UI
  • Selection inspectors with evidence/provenance
  • Performance guards for large datasets
Approach

Research grounded in clarity, provenance, and reproducibility

We build graph + map prototypes with explicit provenance: what is observed, what is inferred, and what is uncertain. The goal is decision support and research iteration.