Hyunwoo Sohn Email & Phone Number

California, United States

Mountain View, California, United States

Columbia, Missouri, United States

• Designed explainable time-series forecasting model integrating diverse social media data, using multi-modal LLMs and LSTM to track religious extremism and domestic terrorism (4 federal grant proposals submitted as co-PI)
• Developed LLM-based text augmentation algorithm, reducing training data needs by 60% and outperforming GPT3.5-based model by 5% accuracy, via constrained worst-case data transformations
• Crafting time-invariant continual learning system, analyzing concept drift in the history of human rights
• Co-instructed Data Science and Social Sciences course for 15+ students, providing lectures and guidance

Boston, Massachusetts, United States

• Alexa AI - Entertainment Domain
• Investigated defect patterns in large-scale utterances (0.3M automotive-related, 50M long-tail), collaborating cross-functionally to propose ML solutions and estimate opportunity costs
• Led design of cohortification algorithms for personalized query rewrites, devising offline testing strategies with proof of concept implementation

Raleigh-Durham, North Carolina Area

• Developed robust disease prediction framework, integrating EHR foundation model with data augmentation [IJCAI24]
• Devised interpretable representation learning framework for multivariate EHRs, modeling disease progression with temporal abstraction, embeddings, and CNN/RNN [IEEE BigData20]
• Designed multi-party disruptive talk detection model for collaborative learning, integrating user-aware algorithms with BERT and attention-based regularization [LAK20 & SIGDial22]
• Implemented topic modeling and information retrieval on diabetes forum texts to enhance care coordination

Raleigh, North Carolina, United States

• Supported teaching of 5 CS courses over 7 semesters, including ML and algorithms, for 500+ students
• Courses: C and Software Tools (Fall '17, Summer '18), Design and Analysis of Algorithms (Spring/Fall '18), Automated Learning and Data Analysis (Fall '19), Advanced Data Structure (Spring '19), AI Academy/Machine.

Raleigh, North Carolina, United States

• Improved NER F1-score in low-resource settings by 34% with model-based data augmentation algorithm leveraging foundation models such as BERT [KDD22]

Cambridge, Massachusetts, United States

• Alexa AI - Entertainment Domain
• Developed stability metrics for Alexa NLU system during feature expansion, assessing model internal behavior, developer pain, and customer impact
• Pioneered continual learning techniques, employing data selection (submodular optimization) and regularization (EWC), achieving 46% relative reduction in broken gates and over 30% enhancement in stability measures

Seoul, Korea

• Led cross-functional collaboration, translating business needs to tech solutions
• Orchestrated development of technical solutions: online store (KT shop) and application systems
• Managed 6 medium-to-large projects, including the iPhone reservation system design and online shop integration

Osan, Gyeonggi, South Korea

• Led command and control system development, managing team of four, reducing defects through enhanced deployment processes, and contributing to system development

Doctor Of Philosophy - Phd, Computer Science

Bachelor Of Science - Bs, Computer Science