Subpart

The air permit workbench.

Artificial intelligence can be good for many tasks and bad for others. One thing it is undeniably good at is connecting data points across huge swaths of information.

Well, as a 23-year-old air permit writer, I have come to realize just how much of my job is connecting data in this fragmented, often subjective industry. One must lean heavily on experienced individuals, old permits, federal guidelines, state guidelines, physics, and math to scrape together a permit likely riddled with errors. And now, I understand why operators hire whole teams of people to put together air permit applications because these questions are difficult: do I need an air permit, which kind, how long will it take, and what will be required?

In the past, only specialists could answer these questions sufficiently, but “times they are a-changing.”

Insert “Subpart,” a full-stack, LLM-guided, air-permitting workbench.

Simply tell Subpart what you want to build (a compressor station, backup generator fleet, glycol dehydrator replacement), and the agent will ask for the information it needs. Once uploaded, the agent will take that information and run it through our proprietary, hard-coded deterministic engine and return a draft application package against your real equipment and emissions data. Every number in the produced package is backed up with regulatory citations, emission factors, or equations.

We are building, starting with midstream industries and NSPS OOOOb. If you are an operator, consultant, agency, or potential partner, then reach out – josephkminton@gmail.com.