Ridgeline · Independent build · 2026 —
Fewer than 4% of construction firms keep a dedicated project controls team. On most capital programs nobody with the right expertise reviews the contractor’s monthly schedule update, payment application, or change proposals. Ridgeline sells the review, not the record.
42,114
priced rows checkedInterrogates a contractor’s bid without opening a drawing — internal arithmetic, the dimensional basis the estimator wrote into each line, ratio outliers against the sheet’s own population, and allowances told apart from errors. Run over a live capital project bid: the arithmetic held across every row, and five quantity items surfaced worth putting back to the contractor.
19,633 ft
piping, fully bore-identifiedA machine-learning per-cell OCR pipeline, built after establishing that whole-table extraction fails on engineering drawings. Complete piping take-off across 7,422 rows with every row carrying a bore, plus 440 numbered piles and 196 supports read from a 259-sheet structural package previously judged unreadable.
100 → 15
sheets worth readingFingerprints each drawing sheet’s content independently of its titleblock and revision number, isolating the case revision control misses — content changed, revision unchanged. That is work nobody priced and no contractor can claim for. A 100-sheet reissue reduces to the fifteen sheets that actually changed.
to the minute
contract dates reproducedAn in-house CPM over Primavera P6 exports with multi-calendar float and retained-logic handling, DCMA 14-point quality gates, and delay attribution by reverting one change and re-running the network. Validated against two consecutive updates of a live capital program: both contract dates reproduced exactly, attribution balancing to zero.
9
screens, built firstBuilt first, to replace Power BI — schedule quality, earned value, payment certification, changes and risk, distilling portfolio status into the written monthly review a controls reviewer would produce. Monte Carlo risk simulation and AACE 65R-11 contingency, with findings phrased as questions to put to the contractor rather than assertions.
I spent a decade performing these audits by hand. Ridgeline encodes that judgment — and refuses to let an automated quantity reach a control budget until a human has verified it.
Every reading is graded — priceable, countable, readable, unreadable — and an automated take-off carries a confidence that keeps it out of a control budget until a person has checked it. The quantities are measured from the drawing’s ruling rather than inferred from recognised text, so a row count is defensible without reading a character.
Every screen maps to the AACE Recommended Practice it implements. That is not decoration: in this discipline a number that cannot cite its method is a number that loses an argument.
Python · FastAPI · SQLAlchemy · PostgreSQL · React · TypeScript · PyMuPDF · Tesseract · Monte Carlo · DCMA 14-point · AACE 65R-11
Built solo with AI coding agents under direction — the same posture I would want from any team using them: the agent writes, the engineer decides, and the tests are the argument.