Qian  Zhang

Qian Zhang

My name is pronounced "ch-i-an j-ah-ng"
Postdoctoral Researcher
Computer Science Department
Samueli School of Engineering
University of California, Los Angeles
Email: zhangqianbots-gotta-get-smarter-than-this@its-really-not-that-hardcs.ucla.edu
Google Scholar
CV Research Teaching Diversity

About

I'm a postdoc at UCLA working with Prof. Miryung Kim. I got my PhD degree from The Chinese University of Hong Kong in Dec. 2017, under the supervision of Prof. Qiang Xu. I have mentored five students, including three female students, at both CUHK and UCLA. All of them have published research papers in top-tier conferences.

I am on the academic job market this year!

Research

My research belongs to the exciting cross-disciplinary area Software Engineering for Data and Compute Intensive Systems, an upcoming field with a unique combination of software engineering, heterogeneous computing, data-intensive scalable computing (DISC) systems, and quantum computing. The vision of my research is to lower the programming, testing, and performance barriers of applicatons that are data and compute intensive.

My PhD research focused on heterogeneous computing with energy-efficient ASIC accelerators and I have published at TCAD, DAC, ICCAD, and DATE. My approach bridges the gap between application characteristics and innovative hardware technologies by trading accuracy for better performance and energy efficiency. During my research, I noticed the difficulties in developing applications on emerging computing platforms. Therefore, my most recent research at UCLA has been focusing on improving developer productivity via automated analysis tools in the emerging domain of big data applications and heterogeneous computing and I have published at ICSE and ASE.

  • BigFuzz (ASE2020) adapts fuzz testing to big data analytics, wherein the testing complexity comes from long setup latency and sources other than control flow such as equivalence class cases of dataflow operators. My work is unique in making fuzzing feasible for big data analytics using framework abstraction.
  • HeteroRefactor (ICSE2020) expands the exising automated code refactoring to reduce the human effort of creating an efficient synthesizable FPGA accelerator using HLS. It contains a novel combination of dynamic invariant analysis, automated refactoring, and selective offloading. This work is the first attempt, to our knowledge, for software engineering researchers to address programming difficulties in heterogeneous computing. This work is a part of Intel/NSF CAPA project. The Intel researchers are encouraging us to tech transfer this work to Intel's HLS compiler I++ as the computer architecture is becoming increasingly inclusive of FPGA incorporation.
  • ApproxIt (TCAD2020,DAC2014) is a quality managment framework of heterogeneous computing with apprximaite kernels for iterative methods. To our knowledge, this is the first approximate computing work in the literature that ensures the final output quality of iterative methods at algorithm-level.
  • ApproxANN (DATE2015) is an energy efficient accelerator for neural networks, ranking within the top 1% most cited papers published that year.

Publications

Software Engineering for Data and Compute Intensive Systems

Energy-Efficient Compute Intensive Platforms

Runtime Quality Management

Resilience-Aware Task Scheduling

Professional Service