GridX Civic Intelligence OS

Client: Uplift Capital • Industry: Civic Intelligence / Data Engineering • Completed: June 1, 2026
Python Pydantic Typer Data Engineering Civic Intelligence Geospatial

Challenge

Understanding where a region is actually growing — which corridors are drawing permits, which employers are expanding, which infrastructure and utility projects are moving — means tracking dozens of scattered public sources: city permit portals, RFP notices, council packets, EDC announcements, and ad-hoc research briefs. None of these speak to each other, most are buried in PDFs or single-tenant government web apps, and the signal decays fast. Keeping a durable, inspectable picture of regional economic activity by hand is impractical, and pushing it into an opaque SaaS tool sacrifices the auditability that civic and corridor analysis demands.

Solution

Built gridx — a local-first operating system for economic and corridor intelligence that keeps the entire archive in plain Markdown, JSONL, and CSV so it stays portable, versionable, and human-auditable, with no database or cloud dependency required for the core slice.

  • Ingestion engine that reads research briefs, chat exports, and public-source snippets, extracts structured signals, and writes normalized Markdown copies alongside an append-only signals.jsonl store
  • Typed signal schema (Pydantic) capturing category, geography, value, deadlines, implications, and provenance — currently spanning 3,000+ signals across permits, infrastructure, RFPs, industrial/retail real estate, housing, and employer-expansion activity
  • Gainesville Permit Radar that validates the city’s Tyler EnerGov Civic Access portal, normalizes permit payloads, and ingests the city’s monthly issued-permit reports — extracting text from PDFs that are mislabeled as CSVs — into a unified local record log
  • Entity graph and wiki linking organizations, agencies, corridors, and projects into an Obsidian-style backlinked knowledge layer (1,500+ entities, 20,000+ links), with JSONL as the machine layer and wikilink Markdown as the thinking layer
  • Thesis engine that clusters accumulated evidence into durable, human-reviewed corridor and infrastructure theses and writes compact weekly and thematic intelligence briefs
  • Deterministic, inspectable scoring and dedupe with idempotent re-ingest, source-policy validation, and a full pytest/ruff suite guarding the data contract

Impact

  • Durable, portable archive of regional civic and corridor intelligence that survives tool churn — every record is plain text and diff-able in git
  • One searchable layer across permits, infrastructure, RFPs, employer expansion, and real-estate activity spanning Cooke, Grayson, Denton, and Collin counties plus the broader Texoma corridor
  • Automated permit-portal monitoring replaces manual checking of a single-tenant government web app, with healthchecks that flag source outages rather than inventing data
  • Weekly and thematic briefs turn raw signal flow into structured, reviewable reads on where the region is moving
  • Local-first, human-in-the-loop by design: every thesis and brief is generated as a draft for human review, never an automated decision
  • Reusable signal schema and ingestion pattern that generalizes to new geographies and source types as the archive grows