Research
Methodology + results, when they're ready.
OmniTakeoff Research is an active program covering active-learning recognizers, evidence-grounded AI architecture, and construction-cost economics. Papers will appear here as they clear review — with the actual DOI / arXiv ID and venue.
Catalog
No published papers yet
Honest answer: we have ongoing research efforts but have not yet published peer-reviewed papers. We will list them here as they appear, with the actual DOI / arXiv ID and venue. We will not list speculative submissions or work-in-progress drafts on this page.
Active research areas
What the team is working on
Active-learning recognizers
Per-customer symbol recognition that converges on shop-specific drawing dialects faster than generic pre-trained baselines. Specific accuracy curves are shared with prospective customers under NDA, not published as marketing claims.
Evidence-grounded AI architecture
Reference architecture for AI quantification systems where every output is traceable to its source — applied to construction takeoff but generalizable to other audit-sensitive domains.
Construction-cost economics
Empirical work on estimator time allocation and bid-cycle compression. We collaborate with customer pilots when their data could inform an industry-useful paper.
How we'll work
Three principles for the research we publish
Reproducible
When we do publish, every benchmark will ship with the test set composition, the evaluation protocol, and the scoring code. If we won't show our work, we won't publish.
Customer-grounded
Every claim will be tied back to specific pilot data with named N-counts and full method disclosure — not “we believe” without numbers behind it.
Honest about limits
Where the data is thin or the conclusion is preliminary, we'll say so. We'd rather under-claim than over-claim.
Co-authoring
We co-author with customers when their data warrants it.
If your firm's pilot data could inform a paper that helps the industry, we'll co-author with you (anonymized or named — your call). Co-authors get named credit and full review rights pre-submission.