The accelerator’s first cohort comprised over 75 global experts and focused on patients in France, Italy, and Spain suffering from atopic dermatitis, a condition that causes dry, itchy, and inflamed skin. The team developed an integrated platform to improve healthcare professionals’ (HCPs) and patients’ awareness of the disease and the available treatment options. Since then, the Digital Accelerator has introduced eight products to propel Sanofi’s modernization, reapplying patterns and building blocks to move faster. “We build solutions with business sponsorship and shared goals to deliver products of value to our patients, providers, or even internal users,” says Sghiouer.
These digital solutions, spanning across research and development, clinical, commercial, and manufacturing, address a range of use cases. These include expediting the discovery of new drug targets and enhancing trial efficiency by incorporating real-world evidence and empowering participants to remotely contribute their data and digital biomarkers, gaining insights from regulatory reports using natural language processing, and improving manufacturing efficiency with predictive AI. An example of commercial transformation fueled by the accelerator is its comprehensive marketing suite, deployed in 18 countries, offering a 360-degree view of HCPs, patients, and customers. “Our commercial operations teams have already begun tapping into this new technology’s power to better interact and engage with patients, partners, and practitioners,” says Sghiouer.
To help teams manage complex projects, Sanofi established “pods” that included a product owner, a scrum master, and a diverse team of technical experts. This structure ensured a comprehensive approach to data-driven AI development backed by a robust serverless technology stack on AWS. “We chose serverless technology early because it offers a rapid learning curve,” says Niek Luttikhuizen, head of engineering at Sanofi. “The cost-effectiveness of AWS is a major advantage, as we only pay for what we use.”
To complement its ongoing AI development processes and use machine learning (ML) at scale, Sanofi uses Amazon SageMaker, a service that is used to build, train, and deploy ML models for any use case. Amazon SageMaker provides not only a robust environment for deployment and scaling but also a long-term road map for the accelerator to add in new ML capabilities without needing to start from scratch, migrate, or balance costs with functionality. Amazon SageMaker Studio, a single web-based interface for complete ML development, is used for all ML development steps, such as building, training, and deploying models. “Using Amazon SageMaker, the accelerator has been able to deliver full sets of APIs in less than 24 hours, with a 30-second timeframe for a developer to test it in real production condition—faster than other development environments,” says Jean Burellier, accelerator platform leader at Sanofi.
Preparing data to obtain quality results is critical for analytics and ML. The accelerator’s ML workflows are supported by a robust set of data science tools on AWS for fast, seamless, and secure data ingestion, processing, treatment, delivery, monitoring, and storage. This includes Amazon DynamoDB, a serverless, NoSQL, fully managed database with single-digit millisecond performance at scale. AWS Glue, a serverless data integration service, helps Sanofi discover, prepare, and integrate data, modernizing the extract, transform, load process.
Amazon API Gateway is used to create, maintain, and secure APIs at any scale, facilitating communication between different parts of the stack. Amazon CloudFront, a content delivery network service, helps secure the delivery of data with low latency and high transfer speeds, and Amazon CloudWatch helps observe and monitor resources and applications.
AWS Lambda, a serverless, event-driven compute service, is used to automatically manage the computing resources required for their applications. Amazon Simple Storage Service (Amazon S3)—an object storage service—helps store and protect vast amounts of data for analytics and backup purposes. “When we need a new capability, if it’s available within AWS, we’ll use AWS,” says Luttikhuizen. “We don’t want people to jump between clouds and learn how to use every technology. We want to go fast, so we standardized our way of working.”
The company is also exploring generative AI through the use of Amazon Bedrock, a fully managed service that offers a choice of high-performing foundation models, to provide the next-generation healthcare that patients and customers need. It organized a hackathon to assess Amazon Bedrock environments on more than 10 projects, leading to the successful deployment of three models designed to enhance employee productivity, streamline business processes, and automate content creation for medical-legal reviews.
