Reconstruct AI

The description states that Reconstruct AI is a drawing-first workspace for construction estimation, built as a hackathon project. It allows users to upload a tender PDF and:

• Render every sheet
• Index it into structural elements (footings, columns, beams, rebar schedules, notes)
• Let users ask questions grounded in specific parts of the drawing
• Reconstruct the whole drawing as an interactive 3D model with provenance traced back to original sheets
• Run a five-agent pipeline that:
  • Reads elements
  • Prices them using live ₹ rates
  • Writes the final BOQ into a Google Sheet

The tool is described as being built with:

• Frontend: Vite + React + TypeScript
• Backend: FastAPI
• 3D rendering: Three.js
• Integrations: Google Sheets, OpenAI APIs

Inference The product appears to be a proof-of-concept or early-stage prototype, not a commercial offering.

The author states that the tool was inspired by inefficiencies in how government tenders are bid on — specifically, the time spent manually reading drawings and pricing them. It positions itself as an AI-powered solution to reduce manual effort and avoid guesswork.

Key claims:

• “Reconstruct turns a tender drawing into a costed, source-grounded Bill of Quantities in mins”
• “Reading every sheet, rebuilding it in 3D, and pricing it in live Indian rates with zero invented numbers.”
• “The drawing itself becomes the interface.”

Inference The positioning is focused on efficiency, accuracy, and trustworthiness — especially around avoiding hallucinations or invented data.

The description states that Reconstruct AI targets construction contractors bidding on government tenders, particularly those using PWD/CPWD conventions (Indian government construction standards). It is built for teams already working in Google Sheets and familiar with tender processes.

It also mentions:

• “EPC contracting teams” as a potential pilot group
• “Someone without formal engineering training” can use it

Inference The ICP appears to be construction estimation teams in India, especially those involved in government tenders, who are looking for faster and more accurate BOQ generation.

There is no evidence of a business model or pricing structure in the description. The project is described as a hackathon submission with no mention of monetization, licensing, or customer acquisition.

Inference No commercial model is evident from the self-reported account.

The system uses:

• FastAPI for backend
• React + Vite + TypeScript for frontend
• Three.js for 3D visualization
• OpenAI APIs for LLM-based reasoning
• Sandboxed quantity arithmetic to ensure deterministic outputs
• Google Sheets integration as the delivery surface

Key technical features:

• Multi-agent pipeline with clear separation between LLM and deterministic code
• Provenance tracking from 3D model back to original PDF sheets
• Handling of messy scans (low-res, rotated, inconsistent scaling)
• “Zero invented numbers” constraint enforced via design

Inference The system is built with a focus on reproducibility, trustworthiness, and accuracy, especially in a domain where outputs have financial consequences.

The description states:

• It was submitted to the OpenAI 2026 hackathon
• It works end-to-end from upload to BOQ delivery
• It has been piloted with a “handful of EPC contracting teams”
• It is not yet commercialized or scaled beyond prototype

Inference The product is at an early stage — likely a proof-of-concept or prototype, not yet in production use or generating revenue.

There is no mention of competitors in the description. However, the problem space (construction estimation using AI) aligns with:

• Existing construction tech tools
• AI-powered document processing and estimation platforms
• Tools that integrate with Google Sheets for BOQ management

Inference The competitive landscape is not described, but it likely includes other AI-based or digital construction tools.

• No revenue or customer data: The project is self-reported and lacks evidence of traction.
• Prototype-only status: It’s a hackathon submission with no indication of commercialization.
• Limited scope: Only supports PWD/CPWD conventions; unclear if it scales beyond that.
• Dependency on LLMs: Despite sandboxing, the pipeline includes LLM agents — raising questions about consistency and auditability at scale.
• Unproven adoption: No evidence of real-world usage or feedback from actual users.

1. What is the current status of the product? Is it being used by any contractors or teams?
2. How does the tool handle ambiguity in drawings — especially when dimensions are missing or unclear?
3. What is the accuracy rate of the BOQs generated, and how is that validated?
4. Are there plans to expand beyond PWD/CPWD conventions?
5. Has the team considered integrating with other platforms (e.g., CAD tools, ERP systems)?
6. How does the “zero invented numbers” constraint affect usability or adoption?