Data Inflexion is a software development startup company specializing in creating libraries and tools for real estate and property listing website developers. Their primary focus is on enhancing experience for customers by providing intelligent solutions that help property buyers and renters find properties tailored to their specific needs. Data Inflexion wanted to leverage Generative AI to offer more refined and personalized search results to end users, but lacked in-house expertise to build and implement such capabilities at scale.
The first of their products, prpt, part of the prpt.io suite is a full featured interactive GPT chat panel to intelligently and dynamically generate listings from user questions.
For example: Can I see a list of properties for a family of four near Dubai Internet City?
prptSLM is a Property and Real estate based Small language Model trained on 3.7 billion parameters with extensive data about the Middle East property scene, including reports, regulations, technical manual, listings, laws etc.
Data Inflexion faced a significant challenge in enhancing the intelligence of property search functionality on real estate websites. Their existing search engine was keyword-based and lacked the capability to understand and process natural language input effectively, which limited its ability to provide personalized, context aware results. With the growing demand for AI powered solutions in the real estate industry, Data Inflexion needed a way to integrate advanced AI capabilities without overhauling their entire infrastructure or hiring specialized AI talent.
Data Inflexion risked falling behind competitors who were already integrating AI-driven solutions, which could erode their market share. Dubai being a global center for safe investments attracts people of all nationalities, and the ability to have users interact with the property websites in languages they are familiar with creates a huge advantage for the customer.
To address Data Inflexion’s challenge, Integra developed a custom solution using Amazon Web Services (AWS) native services and integrated open-source technologies.
The solution, named prpt by Data Inflexion,, was designed to seamlessly integrate with Data Inflexion’s existing real estate websites and provide enhanced property search and filtering capabilities through natural language processing (NLP) and AI-driven intelligence.
At the core of the solution is the integration of Amazon Bedrock, which allows Data Inflexion to leverage Claude Haiku 3.0 as the base AI model for language understanding.These embeddings are stored and managed in Amazon OpenSearch Serverless using Amazon Titan as the embedding model, which serves as the vector database for fast and scalable search performance. The solution also incorporates a retrieval-augmented generation (RAG) approach by enriching the search results with publicly available data sources, further enhancing the system’s filtering and search capabilities.
On the infrastructure side, Integra implemented Amazon ECS Fargate to deploy FastAPI Langchain agents, written in Python, for handling API calls and inference tasks efficiently. AWS CodePipeline was set up for continuous integration and deployment (CI/CD), ensuring that updates to the prpt application could be rolled out quickly and reliably. To manage traffic and API usage, Amazon API Gateway was utilized, with usage plans to throttle API calls and ensure fair distribution of resources.
For security, Nemo Guardrails were deployed to manage security compliance, while Cloudflare was used for web application firewall (WAF) protection and DNS management. Additionally, Amazon CloudWatch was configured for monitoring, logging, and alerting, enabling Data Inflexion to maintain visibility into application performance. Trend Micro Vision One was integrated to enhance the overall security posture of the cloud infrastructure, ensuring data privacy and protection.
Throughout the pre and post-implementation phases, Integra provided ongoing support services, including architecture design, deployment, performance optimization, and security reviews. The solution was also continuously monitored and updated based on feedback and evolving requirements.
The prpt application successfully met Data Inflexion’s goal of enhancing property search functionality through AI powered context aware filtering. By leveraging natural language understanding, the application allows users to express complex queries in plain language, such as family size, preferred locality and amenities.This not only improves the user experience but also increases engagement and the likelihood of conversion.The use of AWS Bedrock helped deliver highly accurate, serverless and scalable AI-driven results. The overall system architecture, powered by ECS Fargate, API Gateway and other AWS services, provides Data Inflexion with a cost-effective and scalable solution that is easy to maintain and update.
Key benefits of the solution included:
Improved user engagement: Personalized search results based on natural language queries resulted in a 10% increase in user interaction with the platform.
Time savings: By automating the filtering process, the solution reduced manual effort for users by up to 20%, enabling them to find relevant properties more quickly.