Bio

I’m currently a senior software engineer in T-Head Semiconductor Corp and focusing on CPU simulator, compiler, and other toolchains targeting RISC-V architecture. Before this, I achieved my Ph.D. degree at Institute of Semiconductors (CAS), during which my research interests included machine vision on edge computing chips, deep neural network compression, and software/hardware co-design.

Internships

• 2021/05 - 2022/06, Huawei 2012 Lab, Technology Research Engineer (intern)
  • In 2012 lab, I worked on the auto-tensorization compiler techniques, aiming at compiling fine-grained user code to tensor-level hardware primitives via polyhedral modeling.
  • Additionally, I also worked on dynamic vision sensor (DVS) data processing. DVS has ultra-high dynamic range and frequency but only responses to motions in the view. These properties make it suitable for automotive scenarios with some dedicated optimizations.
• 2020/10 - 2020/07, SenseTime Research, Researcher (intern)
  • In SenseTime research, I worked on building a flexible deep learning model quantization framework for bridging various quantization algorithms and hardware backends.

Education

• 2017/09 - 2022/06, Institute of Semiconductors (CAS), Ph.D.
  • In ISCAS, I was instructed by Prof. Nanjian Wu who leads a team developing vision chips. The vision chip targets ultra-low power near-sensor visual information processing. I cooperated with my colleagues to deploy convolutional neural networks on the vision chip by low-bitwith quantization and structured pruning.
  • The vision chip I worked on is a SIMD-style image processor controlled by a RISC scalar MPU. All processing instructions are handled by a PE array with 256 PEs. The adjacent pixels covered by a convolution kernel are processed by corresponding PEs. The adjacent PEs (typically 3 to 5) are, of course, have dedicated data paths to fit the accumulation requirement. This data flow is quite different from traditional im2col method as it directly consumes image data spatially without any layout reorder.
• 2013/09 - 2017/07, University of Science and Technology of China, Bachelor
  • In USTC, I majored in Applied Physics and got fundamental trainings on logical thinking and mathematical modelling.