Highlight
Streamlining Documentation Challenges
RNL, a leader in services to the higher education industry, sought to help their clients search and analyse vast quantities of unstructured enterprise documents with Generative AI. They needed a platform to accelerate the development of this complex document processing and retrieval application.
End-to-End Solution with Doubleword
Doubleword enabled rapid development of an intelligent assistant application, moving from concept to proof-of-concept in just one month, accelerating RNL's AI initiatives. The platform's flexibility allowed RNL to develop a tailored high-performance solution deployed within their private environment.
Challenge
Navigating Complex Document Types
RNL's collaboration with higher education institutions often involves managing extensive documentation from various departments, each with unique needs, proprietary reports, and sensitive personal information.
These factors complicated document processing and required careful handling.
While RNL had used traditional ML models in the past to predict student enrolment and success, they aimed to rapidly expand their AI capabilities to provide their clients with state-of-the-art AI powered experiences. They needed a privately deployed, secure solution to support their long-term Generative AI strategy, addressing the concerns of conservative regulated institutions who deal with significant amounts of sensitive data.
Due to regulations to protect personal data, RNL required advanced Retrieval Augmented Generation (RAG) and a privately deployed, air-gapped solution to ensure all data remained securely within their environment. RNL wanted to be sure they could move quickly and build a high quality experience for their clients, so they used Doubleword to rapidly build and deploy self-hosted language and document processing applications.
Solution
Doubleword Enterprise Inference Stack Enhances Secure Higher Education Document Processing
With Doubleword, RNL developed robust RAG and document processing applications within their secure environment. Self-hosting the solution was crucial for RNL, as it allowed them to maintain maximum control over their proprietary documents and ensured the security and privacy of sensitive institutional and student data.
The scalable, self-hosted solution also enabled RNL to increase their user base as needed, without incurring unexpected usage based costs.
Results
Security, Optimized Model Access, and Cost-Efficient Scaling with Doubleword
Using Doubleword's private, self-hosted RAG APIs, RNL could build RNL Answers which enhances information retrieval, content generation and ultimately improves user experience in navigating educational materials.
- Model Independence: The self-hosted solution is heavily optimized for the most popular and best-in-class open-weights models, whilst not being reliant on external APIs.
- Rapid Deployment: From demonstration to prototype within one month, with ongoing support to ensure accelerated time-to-value.
- Scalability: Doubleword powers long-context RAG for search over hundreds of documents while serving thousands of users across different departments.
RNL gained complete control over their AI environment, avoiding dependence on API calls to volatile 3rd-party commercial providers.
The tailored solution assured RNL that they were deploying cutting-edge technology and providing the best experience to their customers. RNL also appreciated the predictability and security of self-hosting coupled with Doubleword's transparent pricing structure.
Future
With Doubleword as the backbone of their Generative AI strategy, RNL plans to expand its use across more AI powered applications for higher education institutions. A key use case will involve leveraging Takeoff to accelerate traditional data processing, reducing tasks from a 4–6 week timeline to just a few days. This will allow RNL teams to focus more on revenue-generating activities.
As Doubleword continues to add new features and functionalities, RNL gains increased efficiency in running applications, providing an even wider scope to build new applications without having to worry about the underlying infrastructure.
