Kartheik Iyer Found a Milky Way-Like Galaxy, and Nicknamed It the Firefly Sparkle

In a new paper out this week in the journal Nature, Kartheik Iyer, a NASA Hubble Fellow at Columbia, along with Lamiya Mowla of Wellesley College and other colleagues, outlines how he and his team used NASA’s James Webb Space Telescope (JWST) to discover a new galaxy that was born in the early years of the universe. Iyer, Mowla, and their colleagues in the CAnadian NIRISS Unbiased Cluster Survey (CANUCS) collaboration, nicknamed the galaxy “Firefly Sparkle,” because of its 10 major clusters of stars that look like fireflies, which shimmer against a backdrop of diffuse light from other, less densely packed stars in the galaxy. To mark the paper’s release, Columbia News caught up with Iyer to discuss the Firefly Sparkle and some of his other projects.

Why is your discovery of the Firefly Sparkle galaxy important?

Telescopes like JWST are essentially time machines. Since light from distant galaxies travels for a long time before it reaches our telescope, we see the most distant objects in the universe as they were when the universe was much younger. The Firefly Sparkle, in particular, shows us something remarkable about how galaxies first formed in our universe. When we look at this galaxy, we're seeing light from when the universe was only 600 million years old, just 5% of its current age. What makes it special is that we can actually see individual massive star clusters within it, like bright cosmic sparkles, thanks to nature's own magnifying glass—gravitational lensing, where a massive galaxy cluster in the foreground bends and magnifies light from the distant Firefly sparkle—combined with JWST's powerful resolution.

These star clusters are incredibly dense and massive, suggesting stars in the early universe sometimes formed in extreme, concentrated bursts in environments quite different from how most stars form today. It's like seeing a galaxy's building blocks being assembled. The galaxy itself is similar in mass to what we think our own Milky Way looked like at that age, giving us a glimpse of how galaxies like ours began their journey.

Understanding these early moments of galaxy formation helps us piece together our cosmic story: how the universe went from its simple beginnings to the rich complexity we see today. This discovery shows us one of the ways the first galaxies were built, star cluster by star cluster.

On another note: What does your new AI technology, Pathfinder, aim to do?

I’ve been interested in machine learning for a long time, especially in thinking about ways to harness its potential to democratize access to information and accelerate scientific discovery. In particular, I was looking for ways to use machine learning to stay on top of all the astronomy research that’s coming out at such a rapid clip, which accelerated even more after the successful launch of JWST. I was at a conference discussing this problem with astronomy colleagues and we started to brainstorm ways to catalog astronomy papers in a ChatGPT-like system. That conversation led us to found a collaboration (UniverseTBD) and eventually develop Pathfinder. It’s a large language model (LLM)—assisted catalog for astronomy papers—papers ingested by Pathfinder are converted into strings of numbers and organized into a landscape (see the map below). When users ask it something, the question is likewise converted into a ‘location’ and mapped onto the landscape, helping it find the right set of papers to form the basis for its answers. It means that the technology can answer a lot of astronomy questions, can stay on top of research, and, since it’s only trained on a finite set of hard data, it won’t hallucinate; every statement it makes is factually backed by scientific research. The best part? It can use the natively multilingual capabilities of modern LLMs to translate its answers based on current research into a variety of languages, lowering the barrier for researchers to reach audiences beyond traditional English-speaking settings.

We’ve trained this system on astronomy papers, but in theory it could be expanded to medicine, biology, and basically any other field that has a public corpus of papers.

What does the Pathfinder map (below) show?

This shows the landscape of 380,000 papers that Pathfinder has trained on. It organizes them by topic and creates a heat map that shows what most papers are about. The mountains are the oldest, most established areas in astronomical research, which have tons written about them. The topic of the sun, heliophysics, is the highest peak, which makes sense since the sun is extensively studied in astronomy. The flatlands tend to be areas of active research, and shallow water represents the really new topics. Deep water is an area where there’s absolutely nothing.

What brought you to astronomy, and to Columbia?

While I’ve always been interested in physics and complex systems, I find astrophysics, and the study of galaxies in particular, to be an utterly fascinating topic. This is mostly due to the fact that studying galaxies brings together many different areas of physics—from nuclear physics in understanding how stars form and evolve, to thermodynamics, fluid dynamics, magnetism and gravity in understanding galaxies as interacting systems, to optics in studying the effects of gravitational lensing, to statistical mechanics in studying the behavior of galaxies as a population. The other factor is that astronomy is a deeply collaborative science, with a diverse community of wonderful people working together to solve big problems.

I have always found the New York City Astronomy community to be a vibrant, dynamic environment, and have had the chance to work with people from Columbia, NYU, and the Flatiron Institute since my PhD days at Rutgers. While I started out as an observational astronomer, I wanted to learn more about the modeling of galaxies in cosmological simulations during my current postdoc, and Columbia was the best place for that. My current advisor, Greg Bryan, and a few collaborators lead the Learning the Universe collaboration, which has a heavy focus on the kinds of problems I am interested in working on.

Do you have any hobbies or favorite activities in New York City? 

As a kid from Mumbai, New York feels like home because of the coastal vibe, the public transport, the street food, and the incredible culture. When I’m not working (which unfortunately isn’t all that much), I’m often just walking and exploring the city; I keep finding new things even now. I also try to catch some live music when I can, or go to improv shows and mixers. I’m also a big fan of reading, and go hunting for books all over the city any chance I get (and highly recommend the yearly Brooklyn bookstore crawl).

Your website mentions that you’re into potatoes, and your email address hints that you are, too. What is it about potatoes?

I don’t have a good explanation for why I like potatoes, and I always have. I like cooking and can make 200 or more potato-based dishes. Potatoes are versatile. Potatoes are kind. Potatoes are friendly. Be a potato.