Yen-Shuo (Eric) Su

Building Foundation Models for Behavior Understanding

PhD researcher at Georgia Tech advancing the frontier of machine learning through foundation models, generative AI, and representation learning. Passionate about research that is meaningful, impactful, and influential.

Yen-Shuo (Eric) Su

About Me

Driven by curiosity, powered by innovation

🎯 My Mission

I focus on research that is meaningful, impactful, and influential. My work spans foundation models, generative AI, and representation learning, with the goal of advancing machine learning to solve complex real-world problems. I believe research thrives on the integration of diverse knowledge and insights from multiple perspectives.

🚀 Research Philosophy

Throughout my academic journey, I've explored reinforcement learning, generative AI, and foundation models. As a machine learning intern at LINE, I worked on large language models, computer vision, and customer targeting. These diverse experiences enable me to approach complex problems from multiple angles and develop innovative solutions.

🌟 Beyond Research

Sharing knowledge is essential to fostering innovation. As advisor of the FinTech & RegTech Club at NTU, I developed workshops and delivered lectures on machine learning, data mining, and smart contracts, reaching over 100 students. I'm committed to empowering others and fostering growth in the ML community.

💡 Looking Forward

I'm always excited to collaborate, learn, and grow. My goal is to shape the future of machine learning and make significant impact in artificial intelligence. Whether it's pushing the boundaries of foundation models or mentoring the next generation of researchers, I'm driven by the potential to create meaningful change.

Research Experience

Advancing AI through innovative research and practical applications

🧠

Behavior Foundation Models

Georgia Institute of Technology - BrainML Lab

August 2024 - Present

  • Developed sliding window self-supervision objective achieving 1.27x speedup in transformer pre-training for time series
  • Designed novel patching architecture combining FC and CNN, reducing reconstruction loss by 24% on MABe dataset
  • Built supervised VAE improving classification accuracy by 18% on DANDI dataset
🎭

3D Generative AI & Vision

Carnegie Mellon University - CyLab

May 2023 - May 2024

  • Engineered 3D-aware facial editing framework reducing FID by 57% over StyleCLIP and InterfaceGAN
  • Optimized EG3D for 8x inference speedup using multi-GPU pipeline
  • Boosted production face-recognition accuracy by 1.82% with 100k synthetic images
🤖

Multi-Agent Reinforcement Learning

National Taiwan University - AI & Blockchain Lab

February 2021 - January 2022

  • Studied effects of relaxing QMIX monotonicity constraint in cooperative settings
  • Experimented with noise disturbance in centralized learning phase of MAPPO
  • Contributed to blockchain consensus research published at IEEE BRAINS 2021
💬

NLP & Computer Vision

National Taiwan University - ML Lab

July 2018 - January 2019

  • Improved video caption generation through attention mechanisms and beam search
  • Trained reinforcement learning agents using policy gradient and TRPO optimization
  • Advised by Dr. Hung-Yi Lee, exploring multimodal learning approaches
💼

Industry ML Applications

LINE Taiwan - Data Engineering Team

July 2021 - April 2022

  • Built visual search engine for LINE stickers, reducing search time by 95%
  • Applied focal loss with XGBoost, doubling hit rates from 6% to 12% in targeted advertising
  • Achieved 20% improvement in ROUGE scores for Chinese paraphrasing with GPT-2
  • 🏆 Placed 2nd in LINE Taiwan AI Hackathon
🔒

Hardware Security & Cryptography

National Taiwan University - IC Design Lab

February 2020 - July 2020

  • Researched elliptic curve cryptographic processors with power analysis resistance
  • Optimized hardware speed and power consumption through heterogeneous architecture
  • Advised by Dr. Tsung-Te Liu, implementing algorithms in MATLAB and Verilog

Publications

Contributing to the advancement of AI research

A Reference-Based 3D Semantic-Aware Framework for Accurate Local Facial Attribute Editing

YK Huang, Y Zheng, YS Su, A Bolimera, H Zhang, F Chen, M Savvides

IJCB 2024
📄 Paper

View-Based Federated Byzantine Agreement System for Environmental Blockchain

YC Liao, YS Su, HC Tsai, KY Chiang, SA Harding, MF Sie, S Liao

IEEE BRAINS 2021
📄 Paper
View Full Publication List on Google Scholar

Technical Expertise

Tools and technologies I work with

💻 Languages

Python C/C++ Java JavaScript Shell SQL MATLAB

🤖 ML Frameworks

PyTorch TensorFlow Hugging Face scikit-learn OpenCV NumPy

☁️ Infrastructure

AWS Docker Kubernetes CUDA PyTorch Distributed Ray

🧠 Core Expertise

Foundation Models Generative AI Self-Supervised Learning 3D Vision Representation Learning Computer Vision NLP Reinforcement Learning

Education

Academic journey and achievements

Georgia Institute of Technology

Doctor of Philosophy (PhD) in Machine Learning

August 2024 - August 2028 (Expected)

Research Focus: Foundation Models, Representation Learning, Behavior Understanding

Advisor: Dr. Anqi Wu

Lab: BrainML

Carnegie Mellon University

Master of Science (MS) in Electrical and Computer Engineering

May 2022 - May 2024

GPA: 3.90/4.0

Key Courses: Multimodal ML, Learning for 3D Vision, On-Device ML, Embedded Deep Learning, Computer Vision, Machine Learning for Signal Processing, Cloud Computing, Parallel Programming

Teaching: Deep Generative Modeling (2024), Embedded Deep Learning (2023)

National Taiwan University

Bachelor of Science (BS) in Electrical Engineering

September 2016 - January 2022

GPA: 3.95/4.0

Key Courses: Deep Learning for NLP, Artificial Intelligence, FinTech & Machine Learning, Blockchain & Big Data, Scientific Computing, Web Programming

Leadership: President of FinTech Club (2021-2022), organized seminars with 200+ participants

National Taiwan University

Bachelor of Business Administration (BBA) in Accounting (Double Major)

September 2019 - January 2022

Interdisciplinary study combining technical and business knowledge

Teaching: Introduction to Decentralized Finance (2022)

Taipei Municipal Chien Kuo High School

Gifted Class of Math and Science

August 2013 - June 2016

Advanced mathematics and science curriculum

Selected Projects

Innovative solutions to challenging problems

Cache-Aware OS Task Scheduler

January 2024 - May 2024

  • Implemented cache hierarchy-aware scheduler for optimized task distribution
  • Achieved 2x speedup on diverse workload benchmarks vs static scheduling
Cache-Aware Scheduler Performance

Cache Hierarchy-Based Task Scheduling

EmoSense: Speech Emotion Recognition

August 2023 - December 2023

  • Compared text-based vs audio-based emotion recognition models
  • Discovered audio models show superior pruning resilience
  • Identified pruning disproportionately affects minority class accuracy
💻 View Code on GitHub

FocusFusion: Attentive Diffusion Model

February 2023 - May 2023

  • Built diffusion model from scratch, tested on Oxford Flowers and Pokemon datasets
  • Integrated attention mechanism improving FID by 12%
  • Reduced noise artifacts in generated images
💻 View Code on GitHub

Web Question Answering with CLIP

February 2023 - May 2023

  • Fine-tuned CLIP for image source retrieval from question-image pairs
  • Improved Recall by 37% over BERT and 30% over VLC

Address Girl: Crypto Wallet Art

March 2022 - June 2022

  • Built StyleGAN pipeline generating unique images from crypto wallet addresses
  • 🏆 Won 1st Place & Most Creative Award in NTU FinTech Competition (60+ teams)

Blind Source Separation for EMG

August 2023 - December 2023

  • Developed pipeline extracting Motor Unit waveforms from HD-EMG time series
  • Designed methodology minimizing reconstruction-original signal disparity

Let's Connect

I'm always open to discussing research collaborations, opportunities, or just chatting about AI!