About Vertical AI
We build AI engineering agents that read code-aligned rules, run real calculations, and emit auditable results — for the engineers and operators that keep the grid running.
Vertical AI LLC builds AI engineering automation for the utility distribution industry. Our flagship product is a pair of agents — one for overhead, one for underground — built on a shared Universal Pole Model and a parallel Universal Vault Model.
We started this project because the same workflow keeps repeating across every utility we've worked with: a PLA file gets exported, manually massaged in a spreadsheet, run through one analysis tool, then another, then another, and finally re-imported. The engineering work itself is sound — but the coordination tax is enormous, and context gets lost at every handoff.
Our answer is to keep one normalized engineering model alive across the entire workflow. Parsing, classification, loading, clearance, sag, span clearance, proposals, and export all read and write the same shape. An AI agent on top can drive the whole thing in plain language, but the underlying services are ordinary HTTP APIs you can call directly from a Python script, a Cloud Function, or a CI pipeline.
We ship to production weekly. The agents are billed per session and per-utility tenanted. The standards corpus is per-utility too — your construction bulletins, your tables, your rules.
Tools that read code-aligned rules, run real calculations, and emit auditable results. The conversational layer is a workflow shell — the math is deterministic.
One Universal Pole Model and one Universal Vault Model. Every tool reads and writes the same shape so context survives across parsing, classification, analysis, proposals, and export.
Analyzers emit structured fix proposals with stable IDs, dependency cascades, and an approve / dismiss lifecycle. Engineers stay in the chair.
GO-95, NESC, RUS, and per-utility construction standards are the source of truth. Automation accelerates engineering — it doesn't replace it.
We work with utility engineering teams, joint-use coordinators, and software platforms that touch overhead or underground distribution. Tell us what you're trying to build.