We spent 15 years doing it the hard way. Now we're automating it.

Architecture in India is stuck in 2005.

An architect designing a 3BHK flat spends 2-3 weeks on documentation alone — AutoCAD plans, elevations, sections, MEP layouts, cost estimation in Excel. The design takes days. The paperwork takes weeks.

For interior designers, clients want photorealistic renders before committing. That means hours in 3ds Max, manual material selection, and manufacturing drawings that take a carpenter 3 days to interpret.

For builders, a complete documentation set for one residential project costs ₹2-3 lakhs and takes 3-4 weeks. At 50 projects a year, that's over ₹1 crore on documentation alone — before a single brick is laid.

One upload. Everything generated.

NirmaanX takes a floor plan — whether it's a hand-drawn sketch on graph paper, a builder's brochure, a scanned PDF, or a CAD file — and runs it through a multi-stage AI pipeline:

The entire process — from uploading a floor plan photo to receiving a complete, construction-ready documentation package with 12 drawing types and a costed BOQ — takes approximately 90 seconds. What currently requires 2-3 weeks of manual work by an architect, a structural engineer, an MEP consultant, an interior designer, and a quantity surveyor is compressed into a single AI-powered workflow.

India's construction industry runs on paper and Excel.

80% of architectural firms in India are small practices with 2-5 people — they can't afford enterprise tools like Revit or hire dedicated BIM specialists. They work in AutoCAD (or hand-drawn plans), estimate costs in Excel, and coordinate with contractors over WhatsApp.

NirmaanX is purpose-built for this market. The platform speaks the language of Indian construction — Vastu compliance, GDCR regulations, Indian Standard room dimensions, local material specifications (BWR plywood, Asian Paints, Hettich hardware), and pricing in lakhs and crores.

Go-to-market is Gujarat-first — starting with Vadodara, Ahmedabad, and Surat — then scaling to Maharashtra, Rajasthan, and pan-India.

Multi-model AI built for architectural understanding.

NirmaanX uses a multi-model AI architecture — Vision Language Models (VLMs) for floor plan understanding, YOLO and classical computer vision for element detection, and large language models for natural language editing and compliance validation. The digitization pipeline is a seven-stage process: preprocessing, wall detection, room segmentation, text extraction, element classification, dimension validation, and compliance checking.

All processing is API-first for the initial 12-18 months, with self-hosted model inference planned as the platform scales. Export formats include DWG, PDF, GLB (3D), IFC (BIM interoperability), and XLSX (BOQ spreadsheets).