How I Built ListingWriter: An Offline AI Assistant for Nigerian Real Estate Agents

What I Built

I developed ListingWriter, an open-source AI application designed to help Nigerian real estate agents create professional property advertisements more efficiently.

ListingWriter generates three types of marketing content from a single property description:

• Professional real estate listings
• WhatsApp marketing messages
• Instagram captions with hashtags

The application supports English, Nigerian Pidgin, and Hausa (beta).

The goal is to make AI-powered marketing accessible to small real estate businesses, including agents who cannot afford expensive AI subscriptions.

Why I Built It

In Nigeria, real estate agents frequently advertise properties through WhatsApp, Instagram, and other digital platforms.

Writing professional advertisements repeatedly takes time, especially when marketing several properties each day.

I wanted to explore whether open-source AI could solve this problem while remaining affordable and privacy-focused.

How It Works

I developed ListingWriter using:

• Python: Application development
• Ollama: Local AI model execution
• Gemma 3 (1B): AI content generation
• Gradio: Interactive web interface

Users enter information such as property type, location, price, bedrooms, bathrooms, and additional features.

They can then generate marketing content based on those details.

Because the AI runs locally, users do not need a paid cloud AI API for every generation.

Technical Challenges

One challenge was balancing model performance with the limited computing resources available on ordinary laptops.

I selected the lightweight Gemma 3 1B model to make the application practical for computers without powerful graphics processors.

Another challenge is language quality. Nigerian Pidgin and Hausa outputs can sometimes contain errors, so generated content should be reviewed before publication.

What I Learned

Building ListingWriter helped me improve my understanding of:

• Integrating open-weight AI models into practical applications
• Running AI locally using Ollama
• Designing user-friendly interfaces
• Developing technology for Nigerian business needs
• Working with multilingual AI generation

Project Demonstration

Watch my project demonstration:

GitHub Repository

The project source code is available here:

https://github.com/hafeezmt/listingwriter

Future Improvements

My next development goals include:

• Better Hausa and Yoruba support
• Saving generated property advertisements
• Exporting marketing content as images
• Improving content consistency
• Exploring deployment options for real estate agencies

Final Thoughts

ListingWriter demonstrates how lightweight, open-weight AI models can support useful applications for small businesses.

I believe developers can create meaningful technology by focusing on everyday problems within their communities.

Project: ListingWriter

Developer: Jungudo Muhammad Tukur

GitHub: https://github.com/hafeezmt