Rebecca Gibson |
Building a useful generative artificial intelligence-powered application is challenging. Take, for example, the data orchestration necessary to create an app that allows consumers to use an image of a house and ask an AI-powered search engine to find similar looking homes in a given price range, in a specific location.
“The app would need to bring together geospatial data about the house’s location, vectors of the images of the house so the AI can analyse its physical attributes and find similar looking houses, and pricing data,” says Abhinav Mehla, regional vice president of cloud partner programmes at developer data platform provider MongoDB.
In order to provide factually correct and contextually relevant responses, developers would need to create retrieval augmented generation (RAG) pipelines that facilitate access to real-time data from multiple sources.
“RAG offers a secure and streamlined way of augmenting large language models (LLMs) with enterprise data to customise their outputs and provide accurate responses, while reducing hallucinations and data leakage,” says Mehla. “It creates better-informed chatbots and improves search and recommendation engines.”
The easiest, most cost-efficient and secure way for developers to build production-ready generative AI apps and RAG pipelines is to use a solution like MongoDB Atlas on Microsoft Azure, which unifies native vector embeddings with live app data into one fully managed, secure, multi-cloud platform.
“MongoDB Atlas spans all transactional, search and retrieval, in-app analytics, geospatial and streaming workload needs, so developers don’t have to bolt on a separate vector database,” says Mehla. “The LLM can access any type of live data to augment generative AI models with up-to-the-second enterprise truth.”
MongoDB Atlas, Microsoft Azure AI Studio and Microsoft Fabric provide an integrated, scalable and secure platform for organisations to bring their data to services like Power BI and Azure OpenAI. “This allows them to harness the best of AI, machine learning and analytics,” says Mehla. “Developers can also quickly accommodate new app requirements and integrate the latest AI innovations. These capabilities are why respondents to an independent AI survey gave MongoDB Atlas Vector Search the highest net promoter score of any similar solution within months of its release.”
To further accelerate the development and wide-scale deployment of generative AI apps, organisations can use the MongoDB AI Applications Program (MAAP), which provides validated reference architectures optimised for different use cases. It also offers an end-to-end technology stack that integrates securely with solutions from providers like Microsoft.
“MAAP offers customers the technology, partners and expertise they need to tackle any AI use case and quickly achieve a return on investment,” says Mehla. “We’ve vetted the components through hundreds of customer implementations and know they’re secure, scalable and work well together as a complete solution. MAPP is continuously enhanced to help customers create generative AI apps that drive instant value for their teams and customers.”
Implement MongoDB Atlas on Azure at: bit.ly/3AV9ZSk
Discover more insights like this in the Autumn 2024 issue of Technology Record. Don’t miss out – subscribe for free today and get future issues delivered straight to your inbox!