Research

I have several research experience in the following directions:

Tensor-based Hardware-Efficient On-device Training Framework with FPGA In-BRAM Computation

[Sept 2024 - June 2025, UC Santa Barbara] Voluntary Researcher advised by Prof. Zheng Zhang

Keywords: Compute-in-Memory, FPGA In-BRAM Computation, Tensor-Train Decomposition Our goal is to develop a tensor-based hardware-efficient on-device training framework, with the idea of implementing contraction of tensor-train using In-BRAM computation on FPGAs
Main reference: Kabir, Md Arafat. ReMoDeL-FPGA: Reconfigurable Memory-centric Array Processor Architecture for Deep-Learning Acceleration on FPGA. Diss. University of Arkansas, 2024.
Paper: to be finished

[July - Sept 2024, New York University Shanghai] Research Assistant advised by Prof. Zhuocheng Xiao

Keywords: Neuronal Oscillation, Spiking Neural Network, Computational Neuroscience
Designed a scalable spiking neuron network in MATLAB to model a small copy of the mouse visual cortex L2/3, generating multiband neuronal oscillations (gamma, beta, alpha, etc.) using data from the Allen Institute (projection probability, synaptic strength, etc.)
Conducted extensive simulations to explore the systematic relationship between biological parameters and the various types of neuronal oscillations observed in the network
Research Document: How Inhibitory Neural Circuits Govern Multiband Neuronal Oscillations

[Feb 2023 - July 2024, HUST] Five-person Research Team Leader advised by Prof. Chao Wang

Keywords: Neuromorphic Hardware Design, Spiking Neuron Model, Hodgkin-Huxley Neuron, Adaptive Exponential Neuron, CORDIC
Proposed a dual-mode spiking neuron hardware design (Hodgkin-Huxley model & Adaptive Exponential model) using an optimized reconfigurable pipeline based on data flow dependency and Reconfigurable Fast-Convergence CORDIC algorithm
Simulated the proposed dual-mode neuron design in Python and conducted hardware implementation in Vivado, demonstrated that our design achieves up to 4.8×improvement in latency and 3.0× in accuracy
Analyzed the implementation results, summarized the design methods, and wrote paper as co-first author
Led the five-person team, and took charge of project schedule management and task allocations
Paper(TENCON’24): A Low-Latency and High-Accuracy Dual-Mode Neuron Design for Accelerating Neurological Diseases Simulation and Analysis

[July - Aug 2023] Four-person Team Leader

Developed surroundings recognition algorithms in Python, allowing the robot to autonomously identify its environment and execute corresponding actions to overcome obstacles such as single-plank bridges, stairs, and minefields
Designed sequences of servo motor parameters to control the robot’s movements, including forward motion, redirection, and rolling over

[Dec 2023 - Jan 2024] Three-person Team Leader

Developed a image processing system on FPGA, capable of performing image enlargement, edge extraction, erosion, and dilation, with output displayed via HDMI
Designed the image processing system’s overall framework, the system control unit and HDMI display module using Verilog Hardware Description Language