Architecture, understood by machines. Build custom architectural plans for South Asia.
Naqsh e Faryadi is Pakistan's premier AI-powered parametric BIM and architectural layout generation engine. Trained the way architects learn: point, line, plane, space, and human clearances.
Mirza Ghalib wrote: "Naqsh faryadi hai kis ki shokhi-e-tehreer ka, kaghazi hai pairahan har paikar-e-tasveer ka." This opening verse of his divan questions the transience of all physical structures. Our business, Naqsh e Faryadi, derives its name from this philosophy. To draw, design, and build architectural layouts is to write a beautiful, temporary plea of form on the blank paper of physical space.
Why Pakistan and the South Asian Context?
Most AI systems are trained on Western datasets like CubiASA or Matterport3D. They lack understanding of regional architecture: central courtyards (Aangan), heavy brick columns, deep verandahs, guest areas (Baithak), and multi-generational housing layouts designed to handle extreme subtropical climates. Naqsh e Faryadi is compiling the world's first Urdu-annotated, culturally relevant parametric dataset of South Asian architectural plans.
With a global AEC (Architecture, Engineering, and Construction) software market expected to reach $1.4T by 2030, and less than 1% of architectural workflows currently augmented by AI reasoning models, there is an immense opportunity to digitize, index, and automate layout design starting from Karachi, Pakistan.
Computational Spatial AI Features
- Reasoning-First CAD: Natural language prompts compile to valid 2D and 3D architectural models in real-time.
- Two-Brain Architecture: Brain One performs spatial reasoning and regulatory code analysis. Brain Two builds the precise mathematical layout and outputs standard parametric IFC code.
- BIM Compliance Checking: Automated validation checks for standard fire egress clearances, habitable room size minimums, and local Pakistani building codes.
Latest Research Blog
Why Pakistani floor plans don't exist in any AI training dataset — and what we're doing about it
Standard architectural datasets are entirely monocultural, trained on thousands of suburban North American single-family homes or European apartments. They fail to understand how thick brick masonry wall structures, solar shade overhangs, double-height natural wind-scoops, and standard South Asian interior zones interact with intense sub-continental solar angles. We explore how we are building and open-sourcing the first specialized spatial BIM dataset to fix this gap.