Kangqi (Amos) Zhang

Shot in Chicago, IL

I am a first-year Ph.D. student in the Elmore Family School of Electrical and Computer Engineering at Purdue University, advised by Prof. Jinghan Sun. My research lies at the intersection of computer architecture, EDA, and machine learning systems.

Before Purdue, I received my B.S.E. in Computer Engineering (Summa Cum Laude) from the University of Michigan, where I worked with Prof. Nathaniel Bleier in the Crucible Computing Group on stochastic computing for mixed-precision DNN inference, chiplet-based systems, and agile hardware design. I also worked with Prof. Jishen Zhao and Prof. Zhijian Liu on efficient long-context LLM inference. In addition, I hold a B.S. in Mechanical Engineering from Shanghai Jiao Tong University (Global College), where I worked as a research assistant in Prof. Yanming Wang’s group on 3D multi-physics modeling of lithium-ion batteries. I also studied at the University of Auckland as an exchange student.

selected publications

HPCA

Precision As You Need: A Stochastic-Computing Mixture-of-Precision Neural Network Accelerator

Haoran Jin^*, Barry Lyu^*, Kangqi Zhang, Jirong Yang, Soren Walther, Jonathan Postelnik, Qiuyi Ding, Justin Shin, Cheng Chiu, Aaron Xiong, and Nathaniel Bleier

In IEEE International Symposium on High-Performance Computer Architecture (HPCA), 2027
NeurIPS

Precision As You Need: Stochastic Computing Is a Dense Adaptive Quantizer

Haoran Jin^*, Kangqi Zhang^*, Jirong Yang, Barry Lyu, Qiuyi Ding, Ruijie Gao, and Nathaniel Bleier

In Advances in Neural Information Processing Systems (NeurIPS), 2026

Abs arXiv

Matrix multiplications dominate the inference cost of modern transformer-based vision models, yet existing efficiency techniques such as post-training quantization and mixed-precision inference are largely limited to the small set of fixed-width formats (INT4, INT8, BF16, and FP16) supported by conventional accelerators. We revisit stochastic computing (SC) as a way to lift this constraint: viewed as a dense adaptive quantizer, SC controls precision by bit-stream length L rather than a fixed datapath, while each multiplication reduces to a single AND/XNOR gate. We build a GPU library that emulates SC matrix multiplication at scale, exposes stream lengths as first-class kernel arguments, and evaluates SC end-to-end on image classification, object detection and instance segmentation, class-conditional image generation, and visual world-model planning. On top of this substrate, we develop a dynamic per-row mixed-precision policy that assigns stream length per token or group at matched average budget, requires no retraining, and uses the same SC hardware across schedules. Across tasks, SC remains competitive with fixed-format INT quantization at matched bit budgets, while per-row mixed precision helps maintain accuracy at lower average stream lengths. These results provide software-level feasibility evidence that SC can serve as a dense-precision substrate for fine-grained mixed-precision inference on modern vision transformers.
ICCAD

TurnRL: Reinforcement-Learning Driven Turn-Restricted Adaptive Routing for Chiplet-Based System

Kangqi Zhang^*, J. Wang^*, Ronald Dreslinski, and Nathaniel Bleier

In IEEE/ACM International Conference on Computer-Aided Design (ICCAD), 2026
MICRO

Fengshui: Demystifying Chiplet Ecosystem and Bespoke Neural Network Accelerator Codesign

Haoran Jin, Jirong Yang, Zhiheng Zhang, Justin Shin, Barry Lyu, Kangqi Zhang, Yunpeng Liu, and Nathaniel Bleier

In IEEE/ACM International Symposium on Microarchitecture (MICRO), 2026

arXiv