Genentech’s CMG team launches Deepsense, using Generative AI to process unstructured data efficiently. By integrating various formats, the product facilitates faster insights and improved patient interactions while maintaining security and regulatory compliance. Looking ahead, Deepsense aims to fuse structured and unstructured data for enhanced decision-making.
Genentech, part of the Roche Group, is taking a bold step forward in how it handles vast amounts of unstructured data through innovative use of Generative Artificial Intelligence (Gen AI). With a focus on enhancing health outcomes for patients and ensuring equitable access, the Commercial, Medical, and Government Affairs (CMG) team is harnessing different data formats—ranging from PDFs and images to videos and audio files. The challenge has always been the complexity and messiness of unstructured data. It’s like searching for a needle in a haystack—difficult, time-consuming, yet loaded with rich insights just waiting to be uncovered.
Facing the difficulties of processing unstructured data efficiently, Genentech’s CMG team decided enough was enough. Enter their new product, “Deepsense.” Instead of relying on cumbersome manual methods to glean insights from data, which previously took weeks or even months, Deepsense aims to streamline and automate these tasks. It gathers and processes insights from various sources, presenting them in a digestible format. Furthermore, attention to security and regulatory requirements safeguards both patient and customer information as decisions are made faster and more effectively.
How exactly does Deepsense work? Well, it’s built on Amazon Web Services (AWS) technology. The frontend uses a web application that’s secured through Amazon’s S3, while data is stored in a dedicated data lake. There’s a data pipeline involved that extracts and processes the information, followed by machine learning capabilities utilizing Amazon SageMaker. This system helps with everything from topic modeling to sentiment analysis, making it easier for users to find what they need without endless searching.
Users logging into Deepsense can quickly search topics and get insights delivered literally at their fingertips, thanks to a seamless integration with Amazon OpenSearch Service. This innovative application doesn’t just let users search, but also allows them to engage with it conversationally—asking questions or requesting summaries in real-time. The tech behind it, Retrieval Augmented Generation (RAG), really takes it to the next level, lending a cutting-edge feel to the whole user experience.
As the generative AI landscape evolves, Deepsense is equipped to tap into Large Language Models (LLMs) like Anthropic Claude 3 and its successor via Amazon Bedrock. This means more accurate extraction of customer sentiment, identification of trends, and even crafting engaging narratives—essentially freeing up the creative capacity of Genentech’s teams. The resulting insights have the potential to transform how the company approaches campaigns and internal communications, shifting the paradigm toward more nimble content creation.
Of course, security is a high priority here. Following guidelines from the AWS Well-Architected Framework, Deepsense has layers of protection from encryption of data in transit to stringent access controls. This sensitive data is accessible solely to authorized users, minimizing risks of unauthorized access and ensuring compliance with regulatory frameworks.
Genentech’s Deepsense marks a technological leap, revolutionizing how unstructured data is processed in the healthcare landscape. With the agility of Generative AI, it not only enhances patient engagement but opens a wealth of insights that can drive business strategies. The integrated approach to both unstructured and structured data will shape decision-making processes further into the future, ensuring the company remains at the forefront of biotech innovation.
Original Source: aws.amazon.com
