The crack of a bat and the roar of a crowd can instantly capture a baseball fan’s attention, even if the home run is missed. Humans rely heavily on vision, yet we also perceive the world through sound. MIT professor Paris Smaragdis is working to give machines hearing abilities that surpass human capabilities, enabling rapid detection of accidents, drilling failures, and other critical events by listening.
He describes his work as “computer‑vision for hearing,” focusing on sound‑source separation, audio‑augmented reality, and tools that let people create, analyze, and manipulate music.
Born in Greece, Smaragdis started with classical guitar, but a small synthesizer his father brought home sparked a fascination with sound synthesis. Traditional European conservatories offered no path for electronic music, so he enrolled at Berklee College of Music in Boston, where he could explore computers and synthesizers and chose bass as his major to venture into a less‑trodden sonic domain.
At Berklee he collaborated with electronic‑production professor Richard Boulanger to build computer‑based music tools, and partnered with MIT Media Lab’s Barry Vercoe on software synthesizers. He completed his master’s, PhD, and postdoc in Vercoe’s group, gradually shifting from signal‑processing techniques to machine‑learning approaches, convinced that teaching computers to listen required data‑driven models rather than hand‑crafted rules.
After his postdoc he joined Mitsubishi Electric Research Labs, creating AI that scans video soundtracks to identify highlights—e.g., a crowd’s roar indicating a key moment—technology later embedded in personal video recorders, traffic‑monitoring systems, and smartphones. He then moved to Adobe, where he built an audio‑research team that delivered user‑friendly sound‑editing tools for artists and producers.
At the University of Illinois Urbana‑Champaign his group applied neural networks to separate individual sources from audio mixtures, a capability that helped clean archival Beatles recordings for the 2021 documentary series “The Beatles: Get Back,” earning an Emmy for sound editing in 2022.
In 2025 he returned to MIT to tackle interdisciplinary challenges. One research line creates systems that listen to musicians and provide constructive feedback without replacing them. The other develops flexible audio‑processing algorithms that discard fixed notions of sampling rates, frame sizes, or context windows, opening new computational structures for more efficient processing. His team also investigates integrating large language models into speech recognition and audio enhancement, noting that the optimal use of AI for sound remains an open question.
He emphasizes hands‑on mentorship, helping students hear the results of equations. Outside the lab he plays jazz bass, explores Boston’s music scene with his family, and coaches underwater hockey—a niche sport he introduced at Illinois and hopes to revive at MIT.
For Smaragdis, pursuing sounds no one has heard before is a continual test of creative ability.