Win your next pursuit with the projects you have already done.
Corpus by SalemWise is page-cited AI for civil, structural, geotechnical, environmental, surveying and mining firms. It reads your firm’s archive of past projects, then helps you decide bid or pass on a new RFQ, draft the SOQ with every sentence cited to the page, and build the fee. Built for principals and BD leads at consultancies of 20 to 500 people.
Most proposal AI reuses your old proposals. Corpus reads the projects themselves: geotech reports, calc packages, specs, bid tabs and drawings, and every sentence cites the page it came from.
From the live demo: a fictional sample firm built from real public engineering documents.
From RFQ to fee, answered from your own past work.
Bid or pass
Corpus reads a new RFQ, checks it against the work you have actually done and suggests go / no-go scores, each linked to the pages behind it. Fit, relationship and capacity stay your call.
Draft the SOQ
Draft the SOQ from your own relevant projects, every sentence cited to the page. Your team edits a draft instead of starting from a blank page.
Price it and paper it
Build a cost-plus-fixed-fee estimate; the demo follows WSDOT’s method. The fee math is done by code, not by the AI. Then start from a draft contract: a draft, not legal advice.
Who sees what
Restricted projects stay restricted. Access is trimmed by role before search runs, so nobody gets an answer from a file they cannot open.
Ask the archive a plain question. Get the page it came from.
Search is the engine underneath the pursuit tools. Engineers ask in plain words and get answers cited to the document and page, so they can open the source and check it.
If it is not in the archive, Corpus says “Nothing on file” instead of guessing.
Question
What watertightness testing standard and leakage limit apply to the containment concrete structures?
Answer
ACI 350.1, Testing Reinforced Concrete Structures for Watertightness. The maximum leakage allowed in the final 24-hour test is 0.05 percent of the test volume [1]
[1] Source: 2020-EV288-STR-Spec · p. 7 (Evergreen Containment Basin)
Question
What liner thickness did we carry for the Harborview digester dome?
Nothing on file.
No customer logos yet. Here is what you can check today.
The live demo
57 public engineering documents, filed under a fictional sample firm: DOT plan sets, geotech reports, specs, bid tabs and real public RFQs. Ask it anything.
Every page citation
Click a source and the cited page opens with the passage highlighted.
The refusals
Ask about work that is not in the archive and it says “Nothing on file”.
The permission trim
Switch from the engineer to the principal: the restricted project appears only for the person allowed to open it.
The fee math
Done by code, line by line, not by the AI.
The price
$5,000 fixed for the pilot, credited toward setup. You judge it on your own RFQ.
Start with one pursuit. Decide after.
Prove it on a real RFQ before you commit to anything larger. Corpus runs on your own hardware (a local NVIDIA GPU on your network) or ours, on an open-weight model with no outside AI service. On-prem, your files never leave your office.
$5,000
Fixed price · about 2–3 weeks
- We load a copy of a slice of your project PDFs, for example the last 2–3 years
- We run your next real RFQ through Corpus: go/no-go, SOQ draft, fee build
- You judge it on your own pursuit
- No integration work in the pilot
The $5,000 is credited toward full deployment if you continue.
Start a $5,000 pilotfrom $21,000 setup
then $1,500–$2,000/mo
- Deployed on your archive where it lives: we connect to your network drive, SharePoint or ProjectWise as part of setup
- In the browser; Microsoft Teams connected during setup
- Runs on your own hardware or ours
- Live in 6–8 weeks. No per-user fees
What engineering firms ask us before they buy anything.
Microsoft 365 Copilot for Civil Engineering Firms
Read the postDeltek Dela for Civil Engineering Firms
Read the postSharePoint Permissions for AI Search
Read the postBest Glean Alternative for Engineering Firms (2026)
Read the postWe also build AI assistants for HOAs, legal teams, invoices and websites.
Built by software engineers, not marketers.
Yermek Ibray (Founder) and Pavel Dolgov (Co-Founder) are software engineers who built systems at Google, Meta and The New York Times. You talk to them, not a sales team.