MS in Computer Science @ University of Southern California (USC)
Focus: Machine Learning • Computer Vision • Audio AI

I build systems that understand and generate real-world signals — from audio deepfake detection to vision-based perception for autonomous systems.

• Audio Deepfake Detection (SynSFX Dataset)

  • Built a large-scale benchmark (43K+ samples) for non-speech synthetic audio detection
  • Evaluated models like AASIST, RawNet2 under domain shift
  • Explored robustness under unseen generators

• TTS + Audio Generation Systems

  • Working with modern models (AudioLDM, VITS, etc.)
  • Exploring synthetic data generation and evaluation pipelines

• Deep Learning for Perception Systems (Vision + Audio)
• Autonomous Driving / Vehicle Intelligence
• Real-time ML systems & deployment
• Robustness under domain shift (very relevant to your research)

Languages:
Python • C++ • C • Java

ML / DL:
PyTorch • Scikit-learn • CUDA (basic)

Tools & Systems:
Linux • Git • Docker (learning)

A large-scale benchmark for non-speech audio deepfake detection

• 43,382 audio samples (real + synthetic)
• Includes prompt metadata + generation provenance
• Designed for reproducible evaluation under domain shift
• Shows that speech-trained detectors fail on general audio

• Email: 17yuyuncong@gmail.com
• LinkedIn: https://www.linkedin.com/in/jeff-yuncong-yu
• Website: https://yuyuncong.com

I like building systems where software meets the real world — whether it's cars, sound, or sensors.