AI engineering for construction
We engineer the AI systems construction runs on.
construction.live is a specialist AI engineering team for construction tech companies and enterprise contractors. We build the harness, memory, cost and control layers that let AI do long, high-stakes work: drawing takeoffs, estimates, contract review and field operations.
Who we build for
Construction tech companies
You are putting AI into your product.
We engineer the parts that decide whether it works in production: agents that finish long tasks, memory that respects permissions, and unit costs that leave you a margin.
Enterprise contractors
You want AI systems of your own.
We build them with your team, on your data, inside your security and approval structure, and hand over a system you own and can extend.
01 · Harness engineering
A controlled harness for long-horizon work.
A takeoff across a full drawing set, an estimate or a contract review is hundreds of steps, not one prompt. We build harnesses that keep the model on a defined path, so long jobs finish correctly and can be trusted.
- Built for long-horizon tasks: drawing takeoffs, estimates, contract and specification review.
- Full control of the run: defined steps, permitted tools, and validation at every stage.
- Checkpoints and recovery, so hours of work survive a failure instead of starting over.
- Embedded end to end: the office directs field operations, and the field reports straight back into the office systems.
02 · Memory layers
Context that holds across people, projects and the organisation.
AI is only as good as what it remembers, and who it is allowed to remember it for. We design memory in three layers and engineer how information moves between them at the infrastructure level.
- Information moves up only when someone promotes it, and down only to people cleared to see it.
- A private working note never surfaces in another team's estimate. A company-wide rate change reaches every live project.
- Permissions are enforced in the data layer, not in the prompt. Get this wrong and the system either leaks data or starves the AI of context.
03 · Cost engineering
The right model for each step, at a cost that scales.
Model prices differ by orders of magnitude. For an enterprise team, or a product with thousands of users, that difference decides whether AI pays for itself. We measure cost per task and engineer it down.
- Routing: each step goes to the smallest model that does it reliably. Frontier models only where they earn it.
- Smaller specialised models, trained on construction work, for high-volume steps like sheet classification and extraction.
- Caching and batching for repeat work across drawing sets and revisions.
- Cost per takeoff, per estimate and per review tracked, so you can price your product or budget your rollout with confidence.
| Step | Model |
|---|---|
| Sheet classification | Small, fine-tuned |
| Quantity extraction | Small to mid, specialised |
| Estimate assembly | Mid-tier |
| Scope and contract risk reasoning | Frontier |
04 · Human in the loop
AI does the work. People keep the authority.
In construction, who signs off matters as much as what was done. Every task the AI completes is verified, and that verification follows your chain of authority.
- Role-based access: each person sees and approves only what their role allows.
- Layered review: an estimator checks the takeoff, a PM approves the estimate, a director signs off the contract position.
- Clean handovers: work passes from one person to the next with its full context and history attached.
- A complete record of every AI action and every human decision, so the hierarchy of decision-making always holds.
05 · Tooling
Workspaces where AI and people work side by side.
AI should not sit in a chat box next to the real work. We put it inside the tools, running in parallel with your people on the same drawing, the same document and the same record.
- Takeoff Beam: our takeoff tool. The AI marks up the drawings and estimators verify them.
- Our own collaborative document editor, built on Tiptap, where AI and people draft proposals, reports and contract reviews together.
- A field app where crews report by voice, photo or message, straight into the office record.
- Connected to the platforms you already run, so the work stays where your teams are.
Who we are
Rahul Vaishnav.
Ten years lived on mining and construction sites. Deep AI expertise that works as your in-house tech team. Behind us, a team across intelligence, cyber security, engineering and claims.
Deployments
Where this already runs.
Multilingual voice and chat for daily updates on a major international airport. The AI read the crew's messages, called whoever had the gaps, and wrote the daily report.
Daily updates tied to schedule and change orders. Flagged a welding method that differed from the contract, worth more than $100k.
Lessons learned initiative automated for a major GC. From 20% reporting to fully compliant.
A GC on a previously abandoned project got workers up to speed within one month.
Proposal writing reduced from 3 days to 6 hours.
Drawing takeoff. Estimators now spend their time verifying the AI-marked drawings.
A multi-billion dollar hospital, two or three drawing revisions a day, and a $100k-a-day delay risk. Scope change detection went from days to a few hours.
Technical report writer for an energy consultant. Reports drafted with 70% less writing time.
Building AI for construction? Bring us the hardest task on your list.
A working session with our engineers. Bring one workflow, a takeoff, an estimate or a contract review, and we will walk through how we would build its harness, memory, cost model and controls.