Three Takeaways from Pfizer at AWS re:Invent Keynote

Contributed by Authors: Lidia Fonseca, Chief Digital and Technology Officer of Pfizer, and Dan Sheeran, General Manager of Healthcare and Life Sciences at AWS.

Pfizer applies its scientific expertise and global resources to bring vaccines and therapies to people that extend and significantly improve their lives. Last year, Pfizer treated 1.3 billion patients – that’s one out of every six humans.

At AWS’s largest annual event, re:Invent, AWS CEO Adam Selipsky shared the keynote stage with Pfizer’s Chief Digital and Technology Officer, Lidia Fonseca. She spoke about the newest generative AI work underway and some of the companies’ most innovative collaborative achievements over the last year and beyond. See the top takeaways below.

 1. Exploring the generative AI future of life sciences: reallocating time and resources to help patients faster

No technology has gotten more buzz in the past year than generative AI. Working with AWS, Pfizer is already harnessing its power to drive innovation and productivity across 17 use cases. From scientific and medical content generation to manufacturing and more, prototypes using Amazon Bedrock and the AWS AI/ML ecosystem have been spun up in a matter of weeks. Pfizer estimates that some of the priority use cases will see cost savings from $750 million to $1 billion annually.

For example, generative AI will help identify new oncology targets. Today, this is a manual process to aggregate information across data sources. But AI can help to identify and collate relevant data and scientific content from many more sources in a fraction of the time. AI algorithms can then assess trends to generate potential targets and help to more quickly validate them and ultimately improve probability of success.

Additionally, using AWS services, Pfizer quickly deployed VOX, its own Pfizer-certified generative AI platform that allows colleagues to innovate in a secure way with access to the best large language models, including Amazon Bedrock and Amazon SageMaker. Other generative AI applications being explored are to create first drafts of patent applications and generate medical and scientific content, for human review and finalization, saving time so Pfizer can get breakthroughs to patients faster.

AWS’s agile culture is very much in line with Pfizer’s way of working, allowing the companies to quickly iterate on prototypes together. And the variety of large language models available in Amazon Bedrock means Pfizer can select the best tools for its specific use cases. AWS’s modular approach allows Pfizer to plug and play with vendors across the board, a significant accelerator for its overall AI strategy.

2. Unprecedented speed and scale enabled by the cloud: the COVID-19 vaccine journey and beyond

Looking back to when the COVID-19 pandemic hit, AWS and Pfizer leaned in to apply technology in meaningful ways. It took just 269 days from when Pfizer announced plans to co-develop a COVID-19 vaccine with BioNTech to the day the FDA granted emergency use authorization. This process normally takes 8-10 years. And prior to the pandemic, Pfizer produced 220 million vaccine doses across its entire portfolio. That number scaled to 4 billion doses of Comirnaty in 2022. A lot of technological collaboration was needed to make this happen.

First, AWS rapidly provided additional high-performance computing capacity, enabling Pfizer to scale into the tens of thousands of additional cores in the cloud. With additional CPUs, Pfizer could conduct computationally intensive analyses to understand how to manufacture its vaccine candidate. Then when Pfizer needed to submit its data in under a day for FDA filings, AWS augmented compute capacity on demand so Pfizer could move at the speed of science.

Next, with the world waiting for the vaccine, Pfizer needed to manufacture and distribute it as quickly as possible. Pfizer’s industry-first Digital Operations Center empowered Pfizer teams to collaborate across plants, see production status, and resolve issues in real time, yielding a 20% throughput increase—this is at the heart of how Pfizer’s plants operate today. For example, an mRNA prediction algorithm yielded 20,000 more vaccine doses per batch.

Today, Pfizer and AWS apply AI to generate proactive alerts regarding events that could potentially disrupt the supply chain, for example allowing intervention in advance of Hurricane Ian to ensure continuity of critical medicine and vaccine delivery.

3. An enterprise-wide data strategy for global success.

Pfizer is now working toward launching 19 medicines and vaccines within 18 months—a feat that has never been accomplished before, by any company. It’s an incredibly ambitious goal, and strong progress—13 launches under the belt—has already been made.

Digital, data and AI are critical to the success of this goal, and what differentiates Pfizer is how it applies these across the entire company. Pfizer started laying this groundwork years ago, putting in place the necessary elements for technology and AI to flourish at the company: centralizing data, creating standards and platforms to scale globally, cultivating strong digital and AI talent, and building a robust and secure foundation, all to innovate for maximum impact.

In 2021, Pfizer embarked on a massive initiative: going from 10% of core IT in the cloud to 80%. This required the migration of 12,000 applications and databases, and 8,000 servers in 42 weeks—one of the fastest migrations AWS has seen with a company the size of Pfizer.

The migration to AWS saved Pfizer over $47 million annually and helped the company exit three datacenters, reducing 4,700 tons of Co2 emissions, or the equivalent of the energy use by 1,000 homes a year. This enabled innovation at speed and scale. For example, Pfizer was able to reduce the time needed to set up computing capacity from months to hours (60k CPUs in 1 hour), which increased speed of data generation for large drug submissions by 75%.

This built on the success of even earlier work: in 2019 Pfizer and AWS developed an industry-leading Scientific Data Cloud (SDC) that aggregates multi-modal data from hundreds of laboratory instruments. The end-to-end platform simplifies the effort involved in processing, storing, retrieving, reusing, and analyzing scientific data generated by laboratory and manufacturing equipment. The SDC’s open data lake architecture being built on AWS allows for the platform to scale dynamically as data volumes increase and analytical requirements become more complex. Since its delivery, the SDC has empowered Pfizer scientists to conduct easy, customized searches of all historical molecule and compound data in real time versus the weeks or months it would have taken in the previously fragmented environment. This time savings enabled Pfizer teams to accelerate analysis and computational research, taking advantage of AI algorithms to identify and design the most promising new compounds.

Watch a replay of the keynote on YouTube

To learn more about how AWS works with life science organizations visit aws.amazon.com/health/life-sciences