Inspiration

I blew up a jet engine while working a project which cost a lot of money JUST because our data was too confusing.

What it does

MEP Ledger helps teams compare HVAC field measurements against a building’s design requirements. Users upload a design document containing equipment labels and target airflows, then import a spreadsheet of measured readings. AI helps extract design values and proposes matches between equipment labels. Users review uncertain values and matches before the software calculates differences, identifies missing readings, and flags results outside a user-defined tolerance.

How we built it

We organized the web application around the technician’s workflow: upload documents, import readings, review matches, and inspect the comparison report. We separated AI-assisted interpretation from numerical calculations. AI helps interpret documents and suggest equipment matches, while explicit calculations determine the differences between design and measured values. For the hackathon, we focused on importing measurement files rather than connecting directly to instruments. This lets us demonstrate the workflow without requiring an airflow capture hood on site.

Challenges we ran into

One challenge was handling inconsistent equipment labels across documents and spreadsheets. A useful comparison depends on connecting each measurement to the correct piece of equipment. Another was deciding where automation should stop and human review should begin. We designed the workflow to surface uncertainty and let users confirm questionable matches. We also needed to present the results clearly enough that someone could quickly identify what needs attention and understand why it was flagged.

Accomplishments that we're proud of

We turned a broad engineering-data problem into a focused workflow with a concrete use case. We’re especially proud of connecting each comparison to its supporting information and giving users control over review tolerances. The result helps users move from scattered files toward a clear list of items that need attention.

What we learned

We learned that extracting numbers is only part of the problem. Those numbers become useful when they are connected to the right equipment, units, requirements, and source documents. We also learned that engineering software needs to make uncertainty visible. A suggested match that a user can verify is more useful than a confident result they cannot trace.

What's next for MEP Ledger

Our next step is to test the workflow with HVAC technicians using real design documents and field measurements. We want to improve document extraction and equipment matching, support more measurement-file formats, and make reports easier to use during follow-up inspections. Longer term, we plan to explore direct instrument imports and expand into additional mechanical, electrical, and plumbing checks.