Wonjune Kang Email & Phone Number

Cambridge, MA, US

Cambridge, MA

Center for Constructive Communication - Exploring applications of speech-augmented large language models (LLMs). Previous work includes developing a speech anonymization system based on voice conversion methods, using NLP methods to study communication dynamics in spoken conversations, and developing computational tools to aid in their qualitative analysis.

Menlo Park, CA

GenAI Speech Team

Menlo Park, CA

AI Speech, Text-to-Speech (TTS) Team - Developed a multi-task learning model for text normalization, part-of-speech tagging, and homograph disambiguation for text processing in a TTS front-end. Work published in ICASSP 2024.

Cambridge, MA

Computational Prototyping Group - Worked on quantifying the robustness of deep neural networks against adversarial perturbations. Extended a pre-existing framework (CNN-Cert) to compute robustness bounds for top-k predictions in image classification models and applied method to new setting of object detection.

Cambridge, MA

Developed a speaker diarization system for spoken conversation data by fusing deep neural network embeddings for speaker identification (d-vectors) with spatial information from acoustic beamforming. Work published in ICASSP 2020.

Cambridge, MA

Performed frame-based activity recognition in videos by fine-tuning image classification models. Augmented the available dataset and optimized classifier performance given limited resources via hard-negative mining.

London, United Kingdom

Data Science Institute - Developed an interactive, large-scale data visualization system for hidden representations of deep convolutional neural networks (CNNs) that was deployed on Imperial College London’s Data Observatory.

Cambridge, MA

Spoken Language Systems Group - Investigated hidden layer representations of Deep Speech 2, a deep end-to-end automatic speech recognition (ASR) model, for the presence of speaker identity information.

Cambridge, MA

Laboratory for Social Machines - Technical lead for the Learning Relationship Machine, a project aimed at modeling human relationship dynamics and patterns through speech and audio analysis.

New Haven, CT

Department of Radiology and Biomedical Imaging - Researched synthesis methods for a novel kappa opioid agonist positron emission tomography (PET) radiotracer.

Seoul, South Korea

DSP & AI Lab, School of Electrical and Electronic Engineering - Developed a high-speed, low complexity algorithm to process and detect the fiducial points of electrocardiogram (ECG) signals (P, QRS, T waves). Work published in EMBC 2015.

Phd, Media Arts And Sciences

Master Of Science, Media Arts And Sciences

Bachelor Of Science, Electrical Engineering And Computer Science, Minors In Mathematics, Economics

Activities and Societies: Varsity Squash, Tau Beta Pi Honor Society, IEEE/ACM, xFair, SuperUROP, UPOP, Theta Xi Fraternity (President)

Summa Cum Laude, Outstanding Scholar (Valedictorian)

Activities and Societies: 1st Squash, 1st Golf, Math Club