Mapping the galaxies, accelerating scientific discoveries and advancing patient care with AI

Since the 1930s, research productivity has fallen at a rate of about 5.1% per year, according to a paper published in the American Economic Review. This decline is largely due to the fact that time is a finite resource and so much of it is occupied with administrative burdens, complex paperwork and fragmented data systems — longtime inefficiencies that AI is now eliminating.

To reclaim time for mission execution and fulfill government mandates to accelerate research through AI, scientific and medical institutions are shifting from traditional computing models to integrated agentic enterprise ecosystems. As 7 in 10 survey respondents say they expect positive AI impacts in both science and healthcare, the technology is already proving its value in speeding the time to science. 

Whether deploying browser-based supercomputing platforms to map billions of galaxies, leveraging ambient intelligence to continuously monitor patient care protocols, or utilizing real-time data fabrics to intercept billions in federal fraud, advanced AI-powered technologies are radically compressing the timeline from data to discovery. 

Managing the scale of the largest astronomical dataset in history 

For many researchers, data has grown to such a massive scale that figuring out how to capture and store that data is a significant hurdle to harnessing it. Frossie Economou, Program Manager at the NSF-DOE Vera C. Rubin Observatory, noted that while today there are approximately 1 billion known galaxies, the Rubin Observatory’s 10-year mission will likely expand that map to about 20 billion — creating “the biggest, most data-rich movie ever made” of the night sky.

“We had to develop a new kind of database system in order to store the data, all these measurements from all these galaxies,” Economou said at Google Cloud Next. “There's a technical challenge, there's a scientific challenge and there's the additional challenge of trying to do this work with really very limited resources, and this is where a lot of the tools and working smart really makes a huge difference.”

The decade-long mission will culminate in a 500-petabyte dataset — the largest astronomical dataset in history — captured by a 3.2-gigapixel camera the size of a car, the largest digital camera in the world. To manage such a cache, the organization built the Rubin Science Platform on Google Cloud, which also serves as a “supercomputer in a browser” to help astronomers access and analyze the data collected by the Rubin Observatory. By leveraging the accessibility of Google Cloud infrastructure, the observatory is democratizing discovery, providing astronomers with all levels of resources the same capacity for analysis and collaboration. 

Deploying ambient intelligence to protect patient safety

Though the stellar domain of astronomers feels particularly limitless, researchers gathering data on earth face similar challenges of scale. Around 10 years ago, Stanford University began identifying sensors and cameras installed in its hospitals that could be used as sources of data. Ehsan Adeli, Assistant Professor at Stanford University, said they now have about 700,000 hours of video they’re looking to mine for AI-enabled patient safety monitoring. 

This project represents the growing field of ambient intelligence, which uses AI to analyze data collected by sensors in everyday environments such as hospitals and medical centers. Through privacy-preserving computer vision, Adeli aims to use AI to improve an area in which many hospitals are struggling: patient care and monitoring, which require strict protocols and checklists, and where resource-strapped clinical staff are already spread thin.

“Everything is manual, even if they have tech … in most cases it comes with a lot of errors and missing data,” Adeli said. An AI system, however, for which Adeli said the team leveraged Google Cloud, Gemini and other frontier models, can assist hospital staff by monitoring activities and checklists, sending alerts for anything missed. “We needed a computer vision system that can real-time analyze behaviors, movements and their actions, and create those alerts,” Adeli explained. 

Patient privacy continues to be paramount — videos are anonymized and rid of any personally identifiable information (PII) and the system’s algorithms are trained to extract information without identifying individuals. By detecting subtle events that human clinicians may miss, the system provides valuable insights while protecting confidentiality.

Real-time inferencing at the subatomic level

For Sergei Gleyzer, Associate Professor of Physics and Astronomy and Chief Science Officer at the University of Alabama, scientific discovery at the subatomic level hinges upon harnessing massive amounts of data created in infinitesimally small windows of time. Protons collide every 25 nanoseconds resulting in exabytes of data to manage.

“You have to make all the decisions if you’re keeping that data and everything that it contains within a microsecond,” Gleyzer explained. “All the decisions, all the best models, all the most powerful AI you can think of needs to run there and inference under that small latency, and the price for that is that you may be throwing away Nobel prizes in nanoseconds if you make the wrong decision. There's not enough physical disk space on this planet for us to store all the data … and that’s where I think AI has a big role to play.”

Historically, Gleyzer said, humans made astronomical discoveries with their own eyes. But just as those observations were surpassed by increasingly powerful telescopes and other technologies, so will AI transform the field. 

“AI is that type of technology – it’s algorithmic and it’s got a lot of hardware behind it … but the reality is it’s sort of working as a lens that gets you to that discovery sooner,” Gleyzer said. He described supervising a Google Summer of Code project that resulted in the discovery of a new type of exoplanet using machine learning. “That was missed by humans that looked at that same system for decades. Humans are imperfect and there is no shame in those discoveries being first identified by tools like machine learning.”

AI-powered clinical decision-making

From subatomic particles to protein biomarkers, AI-powered analysis is also augmenting clinicians’ work at the bedside. With the goal of improving critical injury care, Eric Elster, Dean of the School of Medicine and Executive Vice President for Medical Affairs at the Uniformed Services University (USU), founded the Surgical Critical Care Initiative (SC2i), which uses Google Cloud and Google AI technology to research and develop biomarker-driven clinical decision support tools. 

In SC2i, “we’ve enrolled 3,000 ICU patients, we have 100 million data elements on Google Cloud in a relational database, and about 100,000 bio specimens,” Elster said. 

One of the first tools from SC2i is a decision support tool known as WounDx, which collects protein biomarkers and runs them through an AI algorithm to help surgeons optimize wound closure timing. This automated guidance gives surgeons data-backed confidence in the decision to close to avoid any unnecessary rounds of debridement, or wound cleaning, which can introduce additional risk. 

WounDx research indicates it will reduce rates of wound dehiscence, a post-operative complication, from 23% to 10%, resulting in decreased pain, less time in the hospital and cost savings of around $60,000 per patient.

Transforming fraud prevention from reactive to proactive with AI

Much like time, funding is another finite resource that can have a major impact on the research and health care space. For the Centers for Medicare and Medicaid Services (CMS), responsible for providing health coverage to more than 160 million people, fraud, waste and abuse of funds diverts critical resources away from services and care.

“We pay out about $1.7 trillion a year — let me give you a little scope on that,” said Patrick Newbold, Chief Information Officer for CMS and Director of the Office of Information Technology. “If we were an economy, we’d be the 17th largest in the world.”

With a mission of that scale, responsible and efficient use of resources is critical. The massive and often disparate nature of CMS, however, has also made it a major target for fraud. The agency aims to stop fraudulent claims before they’re paid out, but that requires the ability to flag, in real-time, claims that require deeper investigation. With those needs in mind, CMS partnered with Google Cloud so its environments could be “built for data analytics and AI.”

“We want to move from pay-and-chase to stop-and-call, and we think in the agentic era, we have the ability to do that efficiently,” Newbold said. 

As CMS works to stop fraud, the agency must also ensure programs are easily accessible and navigable for legitimate users. Jaffry Mohammed is Executive Vice President of Transformation, Delivery and Federal Business at Gainwell Technologies, is helping to speed and scale Medicaid eligibility verification in partnership with Google Public Sector. 

“The mandate is to validate employment status and eligibility for the Medicaid program, wherein [users] have to upload certain documents,” Mohammed said. “We apply tools from Google Cloud and Gemini to ingest the data, to do the quality checks and to make sure that it is seamless and it is easy to use.” 

The platform uses algorithms across relevant data sources to increase confidence in identity verification. Gainwell recently announced the AI-driven matching engine achieves 99.5% accuracy. These advancements result in a faster verification process with fewer manual burdens and bring clarity to the correct pathways for users to complete requirements. 

The organization is also working to streamline overly complex health care processes that create inefficiencies in both time and cost. Through the pharmacy platform it runs for Medicaid, for example, Gainwell is taking a more proactive approach to the drug supply chain using predictive modeling. 

“A major source of inefficiencies are data siloes,” Mohammed said, highlighting siloes between drug orders, authorizations and supply chains. “I think the charge for us as professionals is to see where the inefficiency pockets are and how to break them down.”  

Supporting an AI-empowered workforce

Beyond claims processing, Newbold said AI is also helping the agency connect its three core missions — Medicare, Medicaid and market-based insurance. Bringing together these previously siloed missions, CMS can build a unified data fabric to gain powerful, cross-mission insights. These successes, however, are not just built on delivering the tools but also empowering employees to use them well.

“As we talk about investment needs and capabilities, we cannot forget the other side of the coin: ensuring the workforce is equipped,” Newbold said. “I always believe that, especially in our business, the human must remain in the lead … the human must set the context, must set the objectives, must validate the outcome, so that we can make sure that we're doing this responsibly.”

The benefits of AI investment in research and healthcare span across a wide range of use cases, but ultimately, AI-powered transformation is advancing patient support and care, fueling new discoveries, and accelerating breakthroughs. Whether it’s providing efficient and equitable access in healthcare, calculating the right moment to close a critical wound, or expanding our understanding of the entire universe, AI is freeing researchers and clinicians from administrative burdens to focus on their core missions.

Ready to accelerate your mission impact in the agentic era? Tune into the Best of Next for the Public Sector webinar for a deep dive into Google's latest AI advancements and announcements and what they mean for the public sector.