Hye-jin Shim
I'm a speech and audio AI researcher based in Seattle. Most recently, I organized the WildSpoof Challenge, an ICASSP 2026 Grand Challenge. Previously, I was a postdoctoral researcher at Carnegie Mellon University (LTI) working with Shinji Watanabe, and before that at the University of Eastern Finland with Tomi Kinnunen. I received my Ph.D. in Computer Science and Machine Learning from the University of Seoul, advised by Ha-Jin Yu.
My research spans audio deepfake detection, speaker verification, and speech quality assessment, with a particular interest in data bias and evaluation design. Currently, I'm leading a trustworthiness audit of audio-LLM speech-quality judges — testing whether models that rate speech rely on shortcuts, produce calibrated scores, and treat speaker groups fairly.
I'm interested in keeping speech technology worthy of trust — building systems that know when audio is fake, who is really speaking, and how good speech truly sounds, and designing the evaluations that keep them honest.
I'm looking for new opportunities — let's talk
Experience
Independent Researcher · Seattle
Organized the WildSpoof Challenge — an ICASSP 2026 Grand Challenge advancing in-the-wild TTS and spoofing-aware speaker verification. Currently leading a trustworthiness audit of audio-LLM speech-quality judges: shortcut learning, score calibration, and demographic bias.
Postdoctoral Researcher · Carnegie Mellon University, LTI
Led speech quality assessment for channel profiling; co-organized ASVspoof 5 and designed its primary evaluation metric; mentored audio large-model, language-ID, and spoof-diarization research.
Postdoctoral Researcher · University of Eastern Finland
Led SPEECHFAKES, a national project on data bias in audio deepfake detection; exposed shortcut learning in anti-spoofing classifiers.
Ph.D., Computer Science & ML · University of Seoul
Graph attention networks for an integrated speaker-verification and anti-spoofing system.
Blog
Coming soon
Deep dives into audio deepfake detection, speaker verification, and evaluation — thoughts and findings from ongoing work.
Coming soon
Notes from organizing challenges and serving the speech community — WildSpoof, ASVspoof, and what happens behind the scenes.
Coming soon
Life outside research — places, books, and thoughts along the way.
Let's talk.
Open to research and industry roles in speech, audio, and trustworthy AI — and always happy to discuss benchmarks, deepfakes, and evaluation.
shimhz6.6@gmail.com