WellState — an AI-native operating system for oil & gas

WellState is the proof that Auzi builds clinical-grade software, not brochureware — because we didn't build it for a client. We built it for ourselves, in one of the most unforgiving verticals there is: upstream oil & gas compliance.

It is the founder's own company — a separate venture from Auzi — but it is the single most complete expression of the Auzi method. Every surface of the operating system, applied to one industry, built end to end.

What it actually does

WellState manages Asset Retirement Obligations — the legal liability to plug and abandon a well — plus the regulatory compliance that surrounds them. Operators import their well portfolio, the system computes ASC 410-20-compliant fair value with a full audit trail, tracks state-by-state compliance deadlines, stores the evidence, and layers an AI agent system over all of it.

It is not a CRUD demo. The domain model is genuinely deep: bi-temporal obligation records (business time and system time, append-only for audit), a credit-adjusted discount rate decomposed into its risk-free and credit-spread components, e-signed regulatory forms with GPS and IP forensics, and a polymorphic evidence vault.

One person built a multi-tenant SaaS spanning every surface of the operating system — running on real regulator data, not seed fixtures.

The five surfaces, in one vertical

WellState is the Auzi pattern made concrete. Each surface you'd get in your own build already exists here, built and working:

Real data, at industry scale

The backing database holds a public registry built from Texas RRC bulk data: 1.2 million wells, ~78,000 operators, ~15 million production records — kept in its own schema, separate from tenant data, and linked by API number. WellState ingests what the regulator actually publishes, then isolates each operator's working data behind row-level security. (Stat counts are point-in-time from the build; the tilde marks figures that move as data refreshes.)

AI as infrastructure, not a chatbot

More than thirty domain tools across nine areas, seven built-in agents and a custom Agent Studio run through the Vercel AI Gateway, with a pgvector RAG pipeline over the document vault and role-gated tool access. The AI returns real interface — components, tables, computed numbers — not just text.

Why this matters for your build

If we can model bi-temporal liability accounting and ingest a million-well regulator dataset for oil & gas, your practice's scheduling, recall, billing and patient portal are well within reach — built the same way, owned the same way.