Inside the Flatiron Institute: Where algorithms and inquiry shape modern science

The Flatiron Institute stands out for its bold model: full-time scientists, open-source tools, and a mission to accelerate discovery through computation. Its success across fields from astrophysics to neuroscience reflects the power of that approach.

Tucked into the Flatiron District of Manhattan, the Flatiron Institute is becoming an influential hub for discoveries driven by advanced computing. Since opening its doors in 2016, the institute has grown to roughly 250 staff scientists, who are pushing the frontiers of astrophysics, biology, mathematics, neuroscience, and quantum physics.

Andrew Millis, co-director of the Flatiron Institute’s Center for Computational Quantum Physics and a professor of physics at Columbia University (Credit: Simons Foundation)

“What makes Flatiron unique is our combination of first-rate computation and first-rate theoretical and computational ideas," says Andrew Millis, co-director of the Flatiron Institute's Center for Computational Quantum Physics and a professor of physics at Columbia University.

The Flatiron Institute was the brainchild of Jim and Marilyn Simons. 

Jim, who passed away in 2024, was an award-winning mathematician and hedge fund pioneer. Alongside his partner, Marilyn Simons—an economist with decades of experience in philanthropy—he launched the Simons Foundation to back research that might otherwise go unfunded. “As they started to accrue wealth, they wanted to give back to the world," Millis says. “Since their wealth came from mathematics, they sought to fund math and the sciences."

The Simons Foundation’s vision: building a home for data-driven science

Jim and Marilyn established the Simons Foundation in 1994 to help fund groundbreaking research in everything from number theory to quantum gravity to the underlying causes of autism. Over time, “Jim Simons became intrigued not just in funding research, but helping to directly conduct research," Millis says.

Jim and Marilyn Simons, founders of the Simons Foundation (Credit: Simons Foundation)

In 2012, the Simons convened researchers to brainstorm new directions for their philanthropy. There, mathematician and theoretical physicist Ingrid Daubechies “noted that both computation and big data were growing rapidly in importance in science," Millis says. “She recommended they establish a program along those lines."

Jim Simons was especially drawn to this idea. After he earned a doctorate in math in 1962 at the age of 23, he went on to become a code breaker for the National Security Agency. Later, he went on to head the math department at Stony Brook University in New York, where he won geometry's top prize. He left academia in 1978 to found a company later known as Renaissance Technologies, one of the most profitable investment firms in history, where he helped revolutionize finance by applying mathematical models to the trading of investments.

From Wall Street to research: how Jim Simons’ legacy powers the Flatiron Institute

“Jim really appreciated the power of computation and large data sets, so he wanted to establish an institute to push science in that direction," Millis says.

In 2013 the Simons Center for Data Analysis took its first steps toward this goal, demonstrating how carefully designed algorithms could reveal biological insights such as the genetic basis of malignant cancer. “Jim perceived the work this center did as a great success," Millis says. “So he decided to establish a whole institute to do the work."

The Flatiron Institute is designed to address a mission most academic institutions are not well-equipped to perform. Although researchers worldwide are generating increasingly large datasets, analyzing them can take years of effort and expertise in both programming and big-data analysis. Few scientists possess these skills, and their grants are typically not set up to support these projects. 

What makes the Flatiron Institute different from traditional research institutions?

At most universities, researchers must write grant proposals, juggle teaching loads, and assemble ad‑hoc resources for their work. Flatiron scientists, by contrast, are fully supported by the foundation, freeing them to tackle risky, long‑term research. “All codes from the institute are made freely available to the broader scientific community as open-source software," Millis says. This commitment gives researchers immediate access to new tools, insights and the freedom to adapt and build upon each other's work, which is deeply important to scientific progress.

A closer look at Flatiron’s research centers

When the Flatiron Institute launched, nearly all of the work the Simons Center for Data Analysis performed was in the life sciences, so it became the Flatiron Institute's Center for Computational Biology, its first center, Millis says. Its goal is to develop modeling tools and theoretical methods for examining biological data whose scale and complexity have long resisted analysis. Among its accomplishments are the first explanation of how genetic material flows through the nucleus of a working cell.

After some institute-wide brainstorming sessions throughout 2016, astronomy was identified as another priority field where data analysis faced interesting challenges and exciting opportunities, Millis explains. The institute then launched its Center for Computational Astrophysics to investigate the universe on the scale of planetary systems to the entire cosmos. Recently, scientists there helped create the largest-ever map of the universe's quasars.

A map of quasar locations as seen from Earth, with gaps where our galaxy blocks the view. Quasars with redshifts lie farther away. (Credit: ESA/Gaia/DPAC; Lucy Reading-Ikkanda/Simons Foundation; K. Storey-Fisher et al. 2024)

Then, in 2017, the institute launched its Center for Computational Quantum Physics to help solve an infamously difficult problem in quantum physics, the quantum many-body problem. Predicting the behavior of systems of interacting particles is a problem whose difficulty scales exponentially with the number of particles involved. This makes it extraordinarily challenging to solve, even in modestly sized systems. The aim of the center is to use the solutions it can find to predict the behavior of materials and molecules of scientific and technological interest, such as superconductors. Among its accomplishments are using artificial intelligence to reduce a 100,000-equation quantum physics problem to just four without sacrificing accuracy.

A visualization of a mathematical apparatus used to capture the physics and behavior of electrons moving on a lattice. Each pixel shows an interaction between electrons on a lattice—a system once described by 100,000 equations. Using machine learning, researchers compressed it to just four. (Credit: Domenico Di Sante/Flatiron Institute)

After these first three centers, “we decided that what we really needed was a new center to help provide a common intellectual glue for the institute," Millis says. In 2018, the institute created its Center for Computational Mathematics to explore new mathematical approaches to advance research across multiple disciplines. Mathematicians there have helped create massive datasets for training artificial intelligence models to find common threads across seemingly disparate fields, including astrophysics, biology, fluid dynamics, acoustics, and chemistry.

Further discussions at the institute suggested forming another center on the life sciences. In 2021, the institute opened its Center for Computational Neuroscience to understand how the brain works at a computational level through data analysis. Scientists there have developed ways to make brain-mimicking deep learning models more efficient.

The institute still aims to grow “to keep everything fresh," Millis says. “We established the Initiative for Computational Catalysis, starting as a half-size center. It's sort of chemistry, sort of materials science." This initiative currently has a finite duration of 10 years, after which Flatiron will decide whether to fold it into an existing center, let it continue on its own, or move on to something new. 

The future of science at Flatiron: expansion, flexibility, and open collaboration

In addition to these programs, the Scientific Computing Core handles the institute's supercomputing infrastructure and helps scientists use these compute resources effectively. Researchers at Flatiron typically prototype programs on their own desktop or laptop computers, but techniques that might work on modest computers often need to be adapted to run well on a high-performance computing cluster, “at times 100 times faster than what we could get without them," Millis says. 

Millis sees continual renewal as essential. “All in all, parsing big scientific challenges into problems that are computationally feasible is an important part of what we do," he says. “We are not trying to compete on raw hardware, because Google, Microsoft, or Facebook can deploy orders of magnitude more resources than anyone else. Our advantage is having the right concepts, algorithms, and code; that is where Flatiron adds value." And in uncertain times, he adds, “this strategy gives us flexibility."

Charles Q. Choi is a science reporter who has written for Scientific American, The New York Times, The Washington Post, Science, and Nature, among others. In his spare time, he has had the good fortune of adventuring to every continent. He also holds the rank of yondan in the Toyama-ryu battodo style of Japanese swordsmanship.