Zeyuan (Faradawn) Yang

Zeyuan (Faradawn) Yang.

Researcher in distributed systems and ML.

Education

I obtained my B.S. in Mathematics and M.S. in Computer Science from the University of Chicago. I had the privilege to have Professor Haryadi Gunawi as my advisor. My research focuses on Cloud Storage, Distributed Systems, and Machine Learning for Systems.

Research Experiences

1. "Leveraging AI for Faster Storage Access: a Graph-Neural-Network-Based Prefetcher" [2023] [pdf]

Traditional cache prefetchers struggle with the non-sequential access of modern multi-process applications. We develop a hybrid approach combining machine learning to identify access streams with classical algorithms for processing. This method outperforms both ML and huristic methods by up to 10% on traces such as Microsfot Research Cambridge.

2. "Byte-VAE: Novel Memory-Efficient Image Generation Model." [2023] [pdf]

Vector-Quantize Variational Auto-Encoder (VQ-VAE) suffers from high memory usage due to need of a codebook. We designed an innovative rounding method that eliminates the in-memory codebook, achieving near-zero memory consumption while maintaining identical image generation quality.

3. "Scalability Study of Seagate’s Distributed System." [2022] [pdf]

Seagate Technology's distributed storage faced a latency problem. I used the Analytic and Diagnostic Database subsystem to collect metrics like queue depths and operation latencies. Compiling this data, I created a detailed graph of each component's time consumption, leading to a system upgrade that improved throughput by 20%.

Professional Experiences

Video Creation Hobby

Image 1
Image 2
Image 3
Image 4
Image 5
Image 6

App Creation Hobby

Published 3 apps that obtained a 4.9 rating with 5.6k downloads on App Store. The latest app won $10,000 at 2024 Ai4Science Hackathon.

  • "Latin Garden" -- a language learning app endorsed by students from Beijing Forestry University.
  • "Rolling Monsters" -- a game that reduces stress for college students.
  • "Flowers Don't Die" -- an educational app that uses voice AI to accelerate code learning.
</main> </body> </article>