Contact
We're working with a small number of safety and certification teams. If you'd like to see Adducia on your own analysis, or just want to talk about the problem, reach out. A real person will read it.
Reach us
The fastest way to reach us. The link starts a short template; a couple of lines about your team and what you work on is plenty.
Founder. Connect or message there if that's easier than email. Based in the United States; we can meet on a call wherever you are.
What happens next
FAQ
No. Structured tables are parsed deterministically, row by row, with no model involved. Prose and figures are read by AI, and every extracted artifact carries the verbatim quote it came from, verified against your document, with low-confidence reads flagged for an engineer's review. Every artifact ID in an answer is validated against the graph built from your documents, and the checks, the comparison, and the report are computed with no AI in the path at all. The model is the interface. The graph built from your documents is the truth.
In a pilot, your documents sit in a private workspace on a server we administer, under NDA, used only to build your graph and answer your questions, and deleted when the pilot ends. Export-controlled or classified material (ITAR, EAR, CUI) never comes to us; for that, Adducia runs in your own cloud tenant or on your own hardware. Anything public, including demos and this site, uses synthetic data that is clearly labeled as such.
No. Your documents are used to build your graph and answer your team's questions, not to train models.
FMEA and FMECA (including multi-tab Excel workbooks with their severity, occurrence, and detection rating scales), FHAs and hazard logs, fault trees, reliability block diagrams, requirements, and V&V matrices, as XLSX, CSV, ReqIF (the export format of DOORS and Jama), PDF, or pasted text, including the tables and figures where the real content lives. Workbooks and exports are read row by row with no model involved.
Yes. Load a baseline and a newer revision of the same analysis and Adducia reports what appeared, what disappeared, and what changed value, whether that is a severity that moved, a rate that was revised, or a status that closed. It also shows which findings are new or resolved since the baseline. The comparison is computed deterministically, and a renamed or renumbered item is honestly shown as removed plus added rather than guessed at.
Neither. It sits alongside your requirements and analysis tools and hands the engineer organized, source-linked information. Low-confidence reads are flagged for review, and the safety call always stays with the engineer.
For a pilot, a browser and a signed NDA. You get a private workspace with the sample program already loaded, and you bring your own documents. There is no account to create, no card to enter, and nothing to spend. If your program data is export-controlled or classified, Adducia runs in your own cloud tenant or on your own hardware instead, which is the way your export control people would want it anyway.
The pilot is free. Forty-five days on your own work, with nothing to install, no account to create, and nothing to spend. After that, Adducia is licensed annually per program, where a program means one system's safety corpus and every engineer working it is covered. A team running several programs licenses each one. There's no rate card because the price follows the size of the program, which is part of what a first call works out.
We're an early-stage company working closely with a small number of teams. Right now that means free forty-five-day design partnerships: you use Adducia on real work in your own environment, we talk every couple of weeks, and what your team needs shapes what gets built. The first few teams set the direction. Reach out and we'll see if it fits.