Automating 2m Visuals Distribution for Over 75k Hotels

The Challenge

The project involved overcoming several complex, long-standing challenges. As a long-running system, the platform had grown on a non-scalable architecture with performance bottlenecks that made new development slow and the infrastructure inefficient. Deployment processes were fragile, and system reliability suffered as usage scaled.

At the data level, the platform faced a second major challenge. Duplication of images and inconsistent hotel property mapping across third-party booking platforms. Without standard identifiers or consistent naming conventions, matching visual assets to the correct hotel listings introduced significant operational overhead and accuracy issues, especially when dealing with multiple OTAs with different data structures.

Our Solution

Reshape Digital took a steady, structured approach to modernizing the platform. We began by refactoring the entire database storage and optimizing the data layer, addressing long-standing performance bottlenecks and creating a more modular, maintainable architecture. In parallel, we overhauled the infrastructure, introducing scalable cloud services and efficient resource allocation to support high-throughput demands.

“A new JSON-based API enabled seamless media sync across hotel chains and OTAs.”

To further enhance performance and flexibility, we developed a brand-new JSON-based API layer. This served as the core engine for exposing and consuming visual content, enabling structured media ingestion, transformation, and synchronized distribution across hotel chains and OTA platforms. This approach directly addressed the architectural challenges by decoupling systems, improving developer velocity, and laying the groundwork for a more automated and intelligent platform.

“We replaced manual mapping with a machine learning system that matches hotel properties using real-world signals, not static IDs.”

Since there is no standardized database for hotel properties across distribution platforms, we faced a major challenge in matching properties accurately between systems. After extensive research and collaborative brainstorming with our team, we designed a solution using vector databases and machine learning models. We converted property metadata, such as hotel name, location, address, and geolocation, into numerical vectors, allowing us to measure similarity using distance calculations. This enabled us to match entries across platforms with high accuracy, even when naming or structural differences existed, dramatically reducing manual effort and ensuring greater consistency in content delivery.

“AI-driven image recognition and scoring to improve visual content quality”

Beyond the core improvements, we introduced several advanced features to elevate functionality and performance. A new image cropping and optimization tool was built directly into the platform, allowing users to prepare visuals before distribution without leaving the interface. We implemented AI-powered image recognition to automatically categorize photos and suggest tags, reducing manual input and improving consistency. To encourage better content quality, we developed a scoring system that evaluates how well hotels meet visual content standards, helping chains optimize their listings. We also established SEO-friendly visual asset URLs to improve discoverability and performance across booking platforms. These additions, combined with tighter integrations for hotel chain imports and OTA distribution, made the platform more intelligent, user-friendly, and impactful at scale.

Tech Stack

The platform was built using a modern, modular tech stack designed for performance, scalability, and long-term maintainability. The backend runs on C# and .NET 8, delivering speed and stability across services. The frontend was developed in Angular 16, providing a responsive and dynamic user experience. We restructured the database layer using SQL and integrated a vector database to support advanced machine learning–based hotel property matching. CI/CD pipelines were implemented to support multi-environment deployments, enabling fast, automated releases. Media and metadata are managed through custom JSON-based APIs, while AI services handle image recognition, auto-tagging, and asset scoring.

Impact & Results

• Modernized the entire platform infrastructure, replacing a legacy architecture with scalable, modular services that improved performance and resilience.
• Achieved accurate, scalable hotel property mapping using vector databases and machine learning, eliminating manual matching.
• Reduced deployment time and system downtime by introducing robust CI/CD pipelines and automated multi-environment release workflows.
• Enabled seamless, structured integration with hotel chains and OTAs through different connectivity types by normalizing unstructured data.
• Unified services under a scalable JSON API, enabling seamless integration for both hotel chains and OTA platforms, standardizing how data and media are exchanged across the ecosystem.
• Automated image categorization and tagging using AI, significantly cutting down manual effort and improving visual consistency across listings.
• Enhanced user experience and efficiency with built-in tools for cropping, scoring, and preparing images before distribution, driving faster workflows and better visual standards.

What’s Next

We continue to work closely with the client to expand the platform’s AI capabilities, automating visual content intelligence and enabling even more seamless visual distribution at scale. Our focus is on smarter content recommendations, improved categorization, and predictive quality scoring to further reduce manual work and boost impact. In parallel, we’re also supporting the merger of this system with the broader static content platform we've developed (see related use case), consolidating tools and positioning our partner as the leading provider of hotel media and metadata distribution worldwide.