I work for Qualcomm AI Research, and I work in several interesting and interrelated areas. My primary goal is to make computer vision/machine learning workloads efficient either by designing hardware accelerators or making the models more efficient.

Currently I work on following areas:

  • HW-SW co-design for low power machine learning (full stack optimization of ML from model to hardware level. Full stack optimization: Quantization, model level optimization, graph level optimization, hardware level optimization)
  • Efficient 3D computer vision
  • Emerging AI workloads (LLMs, 3D vision, 3D sparse convolution) on edge devices.
  • Collaborated extensively with Universities (UCSD Visual Computing Center, Cornell and many other universities. You can find some of the projects as a result of these collaborations from here.

In the past, I worked on ML compilers (mainly TVM-based), and I also briefly worked on MLIR.

Previous work

Previously, I was a Principal Software engineer at the 3D Engineering department of Cognex Corporation, and I have worked on following areas:

  • FPGA acceleration of computer vision algorithms
  • End to end design of binary neural network on the FPGA

I contributed to the high speed DSMAX product line of Cognex which can be found here: DSMAX 3D LASER DISPLACEMENT SENSOR

Parallel Programming for FPGAs Book

  • Parallel Programming for FPGAs which I contributed (developed most of labs and demos) is used by many Universities around the world starting from UCSD, Cornell, UT Austin, UC Berkeley,..and many others.
  • This book has nice tutorial/demo page here: Parallel Programming for FPGAs,

Academic life: I completed my PhD in Computer Science from the Computer Science and Engineering department of UCSD. I was a Masters’s student at ICU (Information and Communications University) which is now part of Korea Advanced Institute of Science and Technology. I obtained my Bachelor of Science in Computer Science from the Mongolian University of Science and Technology. Other than my current job, I occasionally teach at the Computer Science and Engineering department of the University of California San Diego as a lecturer and an adjunct professor. A long time ago I was in South Korea where I worked for ETRI as a researcher for a couple of years.

My general interests are applications of hw acceleration of ML (LLMs, Transformers, 3D, deep earning, ML), tinyML, ML-based ISP, and computer vision, deep learning compilers.
I am also interested in computer architecture, parallel programming and ML compilers

CV