Aviad Levis joined the Department of Computer Science as assistant professor in July 2024.
Get to know Aviad Levis, who joined the Department of Computer Science as assistant professor in July 2024.
Levis received his PhD in electrical and computer engineering from the Technion – Israel Institute of Technology in 2020.
Your research lies at the intersection of artificial intelligence and physics, developing computational imaging tools for scientific discovery. What attracted you to this area of research in computer science?
I have always been fascinated by physics and signal processing, and in later stage AI. The first time I encountered the term “computational imaging” was during my undergraduate studies. I attended a talk by Yoav Schechner, who later became my PhD advisor, and I was captivated by how this research field seems like magic. Through a combination of signal processing, machine learning, and physics, computational imaging reveals structures that is otherwise completely hidden in the observations. At the time, over a decade ago, that seemed to me like digital magic. Today I work on combining these aspects to design new algorithms that can reveal, learn, and help us discover new structures within our universe.
Prior to joining U of T, you were a postdoctoral researcher at Caltech where you were part of a multi-institution team that generated the first image of the supermassive black hole Sagittarius A* at the centre of the Milky Way galaxy, and earlier this year you published research on a 3D reconstruction of a flare erupting around this black hole. Tell us more about the significance of this research and its impact on the fields of computer science and astrophysics.
Imaging black holes is incredibly challenging, though, despite being notoriously “black”, supermassive black holes are surrounded by extremely bright gas. So, when we say, “an image of a black hole”, what we mean is, an image of the extremely bright gas has a suspiciously dark circular region in its centre. The challenge is that we are trying to resolve such a compact area: imagine trying to photograph a grain of sand in California, when you are in Toronto!
The first image of Sagittarius A* was a collaborative achievement which not only had remarkable agreement with theoretical predictions but also had a wonderful side effect of capturing the broader public imagination and wonder by drawing attention to STEM and the beauty of the universe around us. Occasionally, telescopes also detect extraordinarily bright light from Sagittarius A*, which we call flares.
We wanted to take data acquired in 2017 and go beyond a 2D image to reveal the 3D structure of how a flare could look like. To do that we embarked on a truly interdisciplinary journey to merge technologies developed in computer vision and graphics for 3D imaging of natural scenes (NeRFs or Neural Radiance Fields) and transform them to an environment where the gravitational field is so immense it curves spacetime itself and causes light to move along curved trajectories called geodesics.
The unique combination of neural radiance fields (AI) and general relativity (physics) revealed a structure in the data that would not be available to us otherwise.
As a computer science professor at U of T, how do you plan to build on that research?
My interest is computational imaging for natural sciences and how the combination of AI and physics can accelerate scientific discoveries. I am coming from Caltech with some momentum in astronomy and astrophysics but over the next few years I am also looking to get back to computational imaging for Earth and climate science (check out CloudCT – imaging clouds through a constellation of nanosatellites). I feel like the intersection of AI and physics is still relatively under explored and there is a whole lot of room for creativity.
What’s one thing you hope students who study or work with you will come away with?
On the macro scale I hope students would develop their curiosity towards the large-scale endeavor of scientific discovery and within that appreciate the key role we have to play as modern computational scientists. On the micro scale, that students would tackle diverse interdisciplinary topics and develop the ability to ask the right questions even in a field that is not within their expertise. There is a quote by Edsger W. Dijkstra that really resonates with me: “Computer science is no more about computers than astronomy is about telescopes, biology is about microscopes or chemistry is about beakers and test tubes.”
What drew you to the Department of Computer Science at the University of Toronto?
Interdisciplinary research is challenging; you are constantly out of your comfort zone. It requires embracing a collaborative mindset and tapping into a sense of community. When I came to visit U of T, I was immediately struck by how vibrant, welcoming and collaborative the DCS community is! Another selling point for me was the opportunity to expand the growing Toronto Computational Imaging Group (TCIG) by joining Kyros Kutulakos and David Lindell. I felt like we had a chance to build something truly unique that would push the boundary of imaging.
What are you looking forward to doing or experiencing in Toronto?
I’ve been vegan for the past eight or so years and am really excited to explore the Toronto food scene!
What do you enjoy doing outside of your work as a computer scientist?
I really love the ocean, either just hanging out with my family on a sandy beach day or doing exciting water sports like surfing, snorkeling and swimming. Now that I moved to Toronto, I should probably generalize my love for the ocean to a love for all bodies of water. I am very excited to explore Toronto which I hear has both hiking in nature and hip coffee places within a few minutes’ drive of each other.