Surakshit Ghar: Student Innovation for Safer and Smarter Homes in Nepal
An Innovative Academic Project by Academia International College
Innovation begins when students look beyond textbooks and ask how the knowledge they are gaining can be applied to real-world challenges.
Surakshit Ghar: A Generative 3D Architecture & Grassroots BIM Co-Pilot for Nepal is an academic project developed by students of Academia International College. The project brings together Artificial Intelligence, 3D visualization, Building Information Modeling (BIM), software engineering, and civil engineering concepts to explore a technology-driven approach to residential construction planning in Nepal.
The project reflects the college's focus on encouraging students to explore emerging technologies and develop solutions around real-world problems.
Addressing a Real Challenge in Nepal
Nepal is situated in a highly seismically vulnerable region, with the 2015 Gorkha Earthquake highlighting the importance of safe and resilient construction. Nepal has established building standards, including NBC 205:2024, but the project identifies a continuing gap between technical building requirements and their implementation at the grassroots level.
In many rural communities, residential construction may rely heavily on local contractors, homeowners, verbal descriptions, and hand-drawn sketches. At the same time, professional architectural software can require specialized knowledge, powerful hardware, and expensive licenses.
The students of Academia International College saw an opportunity to explore how emerging AI technologies could help bridge this gap.
Introducing Surakshit Ghar
Surakshit Ghar is envisioned as a Generative 3D Architecture and Grassroots BIM Co-Pilot designed for Nepal.
The project combines natural language processing, generative AI, spatial modeling, BIM concepts, and WebGL-based 3D visualization. Its proposed system can interpret natural-language housing requirements and transform them into structured spatial information for generating dynamic 3D architectural models.
The concept is particularly relevant to students because it requires knowledge from multiple areas of computing and technology.
Through this project, students are able to explore:
- Artificial Intelligence and Large Language Models
- Machine Learning and model fine-tuning
- Web application development
- 3D graphics and visualization
- Building Information Modeling
- Backend API development
- Data validation
- Construction planning concepts
- Cybersecurity and data privacy
From a Simple Idea to an AI-Powered Concept
One of the central ideas behind Surakshit Ghar is to make architectural planning more accessible through natural language.
Instead of requiring every homeowner or local builder to learn sophisticated CAD software, the proposed system allows requirements to be expressed through text.
For example, a user could describe the number of rooms, approximate budget, roof type, and other requirements. The system then processes the description and generates structured spatial information that can be visualized in 3D.
This approach demonstrates how students can use AI not merely as a conversational tool, but as part of a larger engineering and visualization pipeline.
Four Key Technologies Behind the Project
The students have organized the Surakshit Ghar concept around four major technological pillars.
Text-to-Spatial-JSON Generative Architecture
The project proposes using a fine-tuned open-weights language model as a spatial parser. Natural-language construction requirements are transformed into structured 3D spatial coordinate arrays.
This creates a bridge between human language and computer-generated architectural information.
Budget-Aware Design
Construction planning must consider financial limitations. Surakshit Ghar therefore incorporates the user's budget into its proposed design process.
The system accepts a budget in Nepalese Rupees and works backward using local material information to determine a suitable maximum safe footprint.
Deterministic Safety Validation
The students recognize an important limitation of generative AI: an AI-generated design may appear convincing while still violating physical or regulatory constraints.
To address this, Surakshit Ghar proposes a deterministic validation layer using Python and Pydantic to check generated spatial layouts against NBC 205:2024-related constraints.
This creates a simple but important design philosophy:
AI generates. Engineering rules validate.
Procedural 3D BIM Visualization
After validation, the proposed system uses React Three Fiber and WebGL technologies to generate interactive 3D representations.
The visualization can include columns, beams, internal reinforcement cages, and MEP routing layers. The project also proposes X-Ray visualization to help users inspect internal building components.
A Multidisciplinary Learning Experience
One of the strongest aspects of Surakshit Ghar as an academic project is its multidisciplinary nature.
Students are not working on only a conventional website or an isolated machine-learning model. They are attempting to connect several technologies into one system.
The proposed architecture includes a React/Vite/Tailwind/React Three Fiber client, a Python FastAPI API layer, a fine-tuned Llama-3.1-8B generative intelligence layer, and a Pydantic-based validation layer.
This gives students practical exposure to technologies that are increasingly important in modern software development.
Learning Through the Development Process
The project follows an iterative development approach divided into several phases.
The first phase focuses on dataset curation and mapping construction-related language to spatial JSON structures.
The second phase focuses on fine-tuning Llama-3.1-8B using QLoRA and Unsloth.
The third phase involves developing the Python/Pydantic safety gate for validating AI-generated spatial coordinates.
The fourth phase focuses on building the React Three Fiber WebGL engine for procedural 3D visualization and X-Ray inspection.
Through these phases, students gain experience in the complete development lifecycle—from research and data preparation to AI development, backend engineering, validation, and frontend visualization.
Connecting Academia with Society
Surakshit Ghar demonstrates how an academic project can be connected to a problem beyond the classroom.
The students are exploring how technologies taught and researched in an academic environment can be applied to Nepal's construction context.
The project is particularly focused on low-rise residential reinforced-concrete frame buildings of up to three stories and typical rectangular structures.
Rather than attempting to solve every architectural challenge, the students have defined a specific scope that can serve as a foundation for further research and development.
Responsible Use of AI
The students also understand that AI must be used responsibly when dealing with construction and structural safety.
Surakshit Ghar is intended to function as an engineering co-pilot, not as a replacement for qualified professionals. The project documentation explicitly recognizes the need for final sign-off from licensed municipal structural engineers.
The project also considers data privacy, equitable access, secure data transmission, and regulatory compliance.
This emphasis on responsible innovation is an important part of the students' academic journey.
The Vision of Academia International College
Surakshit Ghar represents more than a technology prototype. It represents the creativity, curiosity, and problem-solving approach that academic projects can encourage.
By working on a real-world problem, students have the opportunity to move beyond theoretical knowledge and explore how emerging technologies can be combined to create meaningful solutions.
The project demonstrates that students can take ideas from AI, software engineering, 3D visualization, and engineering concepts and bring them together around a problem that is relevant to Nepal.
Looking Toward the Future
The current project has defined limitations. The proposed system focuses on typical rectangular reinforced-concrete structures and does not currently support irregular or high-rise buildings. Real-time natural-language generation also depends on internet connectivity, although the PWA is designed to support offline input storage.
These limitations provide opportunities for future student research and development.
Future work could explore expanded building types, improved cost estimation, additional building-code rules, enhanced AI spatial reasoning, richer BIM outputs, and broader testing.
A Student Project with a Bigger Vision
Surakshit Ghar is a reflection of what can happen when students are encouraged to experiment with emerging technologies and apply their knowledge to real-world challenges.
Developed by students of Academia International College, the project brings together AI, BIM, 3D modeling, software engineering, and construction concepts to explore a new approach to residential planning in Nepal.
The project may be an academic initiative today, but the questions it explores are highly relevant to the future:
Can AI make architectural planning more accessible?
Can 3D visualization help communities better understand their homes?
Can deterministic engineering rules make generative systems more responsible?
Can student innovation contribute to safer and smarter construction in Nepal?
Surakshit Ghar is the students' effort to explore these possibilities.
SURAKSHIT GHAR
A Generative 3D Architecture & Grassroots BIM Co-Pilot for Nepal
An Academic Project by the Students of Academia International College
Innovating through technology. Learning through practice. Building ideas for a safer Nepal.
