Groundwater district operations platform

Groundwater conservation districts are regulatory bodies: they permit wells, track pumping against allocations, enforce drought restrictions, and report to the state. Districts like CUWCD, BSEACD, North Plains GCD, Hays Trinity GCD, and Middle Trinity GCD each run on their own configuration of this platform: same core, different rules, different aquifers, different politics.

The data is regulated and audited. The users run from hydrogeologists to well owners filing one meter reading a year. Each district is a separate tenant with its own permissions, workflows and public surface.

The front-end tooling, and much of the full stack, behind the operational workflows that districts use every day:

• —Permit management and allocation tracking: the ledger a district's authority rests on
• —Well registration and searchable well registries with spatial context
• —Telemetry viewers and pumping-vs-water-level analysis with well-construction overlays
• —Batch data import with approval and reconciliation workflows
• —A mobile field-services app for well inspections
• —Self-service pumpage reporting for well owners, and DMR regulatory reporting out the other end
• —Drought stage and compliance tracking

Rarely a single feature. The decisions underneath them:

• —Multi-tenancy that holds up when every tenant is a regulator: scoped RBAC via Auth0, per-district configuration instead of per-district forks
• —Data modeling for regulated records: allocations, meter readings, and amendments that need field-level audit trails rather than soft deletes
• —A batch import must not be able to silently corrupt a compliance ledger: review, reconcile, then commit
• —Time-series telemetry at database scale (materialized views, PostGIS spatial joins) feeding charts a board member can read
• —Interfaces spanning the whole literacy range: hydrogeologist dashboards down to a once-a-year meter-reading form a rancher can finish

Not the way it is now. This platform predates the spec-and-loop workflow, but "before AI" would be the wrong way to describe it. AI has been in this work since roughly 2022, and what changed over those years was how much of it I handed over. What the platform is evidence of is the production judgment underneath all of that, which is the part no workflow supplies.