CV

My curriculum vitae.

General Information

Full Name Yanghonghui Chen (Ryan)
Pronouns He/Him/His
Date of Birth May 2003
Languages English, Chinese

Education

2025.09 -
Master of Science in Electrical and Computer Engineering
The University of California, Los Angeles (UCLA), Los Angeles, US
2021.09 - 2025.06
Bachelor of Science in Electrical Engineering
University of Illinois Urbana-Champaign, Urbana, US
2021.09 - 2025.06
Bachelor of Engineering in Electrical Engineering and Automation
Zhejiang University, Haining, China

Open Source Projects

October 2025 - December 2025
High-Precision Fully-Differential Folded-Cascode Op-Amp Design (180nm CMOS)
Course Project (Advisor-- Prof. Behzad Razavi)
Fully-Differential Folded-Cascode Triode-Region CMFB Low-Power Design Cadence Virtuoso 180nm CMOS
  • Designed a high-precision, fully-differential folded-cascode op-amp in 180 nm CMOS with NMOS input pairs to maximize gm/BW; implemented a continuous-time CMFB using triode-region devices to preserve headroom and support 1.6 Vpp output swing.
  • Engineered a low-power design variant through rigorous power-speed trade-off analysis, achieving 2.21 mW total power (~78% reduction vs. high-speed baseline) while maintaining <1% gain error and 45 ns (99%) settling into a 2 pF load.
  • Resolved a critical settling bottleneck by identifying excessive CMFB-network capacitive loading; reduced triode device dimensions by 50% to cut parasitic capacitance—improving large-signal slewing and small-signal settling—and added CLM-compensation resistors in bias branches to ensure VDS-matched current mirrors.
October 2025 - December 2025
Neural Speech Decoding System (Brain-to-Text)
Course Project (Advisor-- Prof. Jonathan Kao)
PyTorch LSTM/GRU BCI Real-time Decoding CTC Loss Google Cloud Platform (GCP)
  • Engineered a real-time, uni-directional LSTM decoder with a specialized non-linear post-stack (Linear-LayerNorm-GELU) to capture long-range temporal dependencies. Enforced strict causal constraints to ensure viability for live Brain-Computer Interface applications, reducing error rates from 23.6% to 19.59%.
  • Developed a robust training pipeline using Focal CTC Loss to address severe phoneme class imbalance in intracranial neural data. Stabilized convergence on Google Cloud (Tesla T4) through Gradient Clipping, AdamW optimization, and a sequential learning rate scheduler.
October 2025 - December 2025
High-Speed MASH-1-1-1 Delta-Sigma Modulator Design in 16nm FinFET
Course Project (Advisor-- Prof. Hooman Darabi)
Verilog ASIC Design Flow Logic Synthesis Static Timing Analysis (STA) Synopsys DC/PrimeTime TCL
  • Engineered a high-speed, pipelined MASH-1-1-1 digital modulator targeting 500 MHz clock frequency. Implemented robust noise-shaping logic using 5-bit signed arithmetic and bit-extension techniques to handle dynamic ranges without overflow.
  • Synthesized the design using TSMC 16nm FinFET technology with Synopsys Design Compiler. Achieved timing closure across Multi-Mode Multi-Corner (MMMC) conditions by developing custom TCL scripts to automatically resolve critical hold-time violations.
  • Conducted gate-level simulations in ModelSim to generate switching activity (VCD) files. Performed high-precision sign-off power analysis using Synopsys PrimeTime PX, achieving a finalized power consumption of 0.178 mW.
February 2025 - May 2025
Four-Axis Vacuum Stage for Advanced Nano-Manufacturing
Senior Design (Advisor-- Prof. Oleskiy Penkov)
STM32 RS-485 Stepper Motor Control PCB Design HMI FSM C/C++
  • Control System Design: Engineered an STM32-based control system for a 4-DOF robotic arm, using the RS-485 protocol for robust communication with four stepper motor drivers. Developed control logic for precise multi-axis motion and implemented a Finite State Machine (FSM) to automate complex coating sequences.
  • Custom PCB Development: Independently designed and validated a custom PCB to integrate power distribution (24 V) and the RS-485 communication bus. The design minimized signal interference and voltage drop, ensuring system reliability in a high-vacuum environment.
  • HMI & User Interface: Created an intuitive HMI using a TFT touchscreen and physical buttons. The interface provides real-time monitoring of motor status (speed, position) and allows one-touch execution of preset programs and emergency stops.
February 2025 - May 2025
Scalable EEG Signal Classification via Federated Learning
Senior Thesis (Advisor-- Prof. Howard Yang)
TensorFlow Federated Learning Personalized FL (pFedMe) EEG Seizure Prediction
  • Designed and implemented a privacy-preserving pipeline for seizure prediction, processing EEG signals into Pearson Correlation Coefficient (PCC) matrices and training a CNN classifier using four distinct paradigms: Local, Centralized, FedAvg, and pFedMe.
  • Demonstrated that standard FedAvg underperforms due to client data heterogeneity and severe class imbalance, while personalized pFedMe significantly improves average sensitivity and F1-scores by adapting to unique patient data.
  • Conducted a detailed hyperparameter analysis of the pFedMe personalization weight (λ), revealing that smaller λ values are critical for minority-class detection (sensitivity), whereas larger values force adherence to a globally biased model, collapsing performance.
October 2024 - December 2024
Optimized Convolutional Layer Implementation Using CUDA
Course Project (Advisor-- Prof. Volodymyr Kindratenko)
CUDA Parallel programming GPU acceleration Convolutional neural networks (CNNs)
  • Designed and implemented the forward pass of convolutional layers for a modified LeNet-5 architecture using CUDA, optimizing performance for deep learning tasks such as image classification and object detection.
  • Implemented a GPU-based forward convolution with a structured Prolog-Kernel-Epilog approach, ensuring memory management, convolution computation, and output transfer, while matching CPU implementation correctness and optimizing performance using Nsight profiling tools.
  • Applied advanced GPU programming techniques to optimize the implementation, including streams, GEMM kernels, and kernel fusion, to achieve a target inference time of ≤80ms for 10,000 images from the Fashion MNIST dataset.
May 2024 - July 2024
RRAM-based Heterogeneous Processing for Multimodal Brain-Computer Interfaces
The University of Hong Kong (Advisor-- Dr. Zhengwu Liu)
RRAM Multimodal brain-computer interface EEG Compute-in-memory FDT
  • Simulated representative heterogeneous processing paradigm of P300 signal recognition in Python by using resistive random-access memory (RRAM) with and without fixed parameter disturbance training (FDT).
  • Contributed to combining the BCIC IV IIa and the P300 RSVP datasets and designing an RRAM-based multimodal recognizer that integrates components of the pre-trained EEGNet, CSP (Common Spatial Pattern), and a modality-fused classifier to create the multimodal settings.
  • Achieved 2.83% higher accuracy using multimodal BCI with FDT than that without FDT and significantly outperformed the MI-alone and P300-alone results by 8.19% and 13.20%, respectively.
March 2024 - May 2024
Raspberry Pi Based IoT System as a Private Chatbot
Course Project (Advisor-- Prof. Deming Chen)
IoT System Raspberry Pi Machine Learning Deep Learning
  • Developed an IoT system using Raspberry Pi 4 as a private chatbot with face detection and speaker recognition to guarantee privacy and personal conversations as well as interactions.
  • Implemented MTCNN with ResNet and dlib-based face recognition, achieving better performance with the latter; trained the system with one hundred face images for live recognition.
  • Built a custom residual neural network with Keras for speaker recognition, achieving 96% accuracy.
  • Integrated a server-client architecture using Google Cloud for accelerated processing and implemented speech recognition and TTS for user interaction.
March 2024 - May 2024
Multiplayer Action Game on FPGA-- Crazy Arcade
Course Project (Advisor-- Prof. Zuofu Cheng)
SystemVerilog FPGAs System-on-a-chip MicroBlaze CPU VGA
  • Used FPGA for real-time operations, integrating MicroBlaze CPU for game logic and keyboard input processing.
  • Developed various modules in SystemVerilog to manage player movements, bomb mechanics, life counts, and game states, interacting through a system bus.
  • Incorporated background music by PWM for sound generation, featuring distinct tracks for different game stages.
March 2024 - May 2024
Cheat-Machine for Game 2048
Course Project (Advisor-- Prof. Thomas Moon)
Embedded DSP Real-time Signal Processing Image Processing Android Studio
  • Developed an app in Android Studio to analyze a live game of 2048, recognizing board digits using image processing.
  • Employed efficient template matching for multi-digit recognition, using grayscale conversion, Canny edge detection, and perspective transformation to preprocess images. Used python packages to evaluate the workflow of the application.
  • Built an AI engine with an Expecti-max Search algorithm to recommend the optimal move, focusing on corner placement strategies.
  • Achieved high accuracy in digit recognition (100% when properly aligned) and consistent AI performance, reaching 1024 tile in 75% of simulations.
December 2023
Simulating Neuron Circuit Design
Course Project (Advisor-- Prof. Jont Allen)
Neuron Simulation Hodgkin-Huxley Model Circuit Simulation
  • Designed an electronic circuit based on the Hodgkin-Huxley model to simulate neuron action potentials.
  • Conducted electrical pulse stimulation to observe and analyze neuronal signaling pathways.
  • Optimized circuit performance by adjusting component parameters to achieve clear action potential observations.
June 2023 - August 2023
Hook&Hair Structure 3D-Printing based on Path Control and 4D Printing Experiment Exploration
Zhejiang University (Advisor-- Prof. Guanyun Wang)
3D printing Grasshopper Rhino FDM Path-planning 4D printing
  • Developed 3D printing techniques for complex hook and hair structures using Rhino and Grasshopper for path planning, generating G-codes for customized printing paths instead of traditional FDM (Fused Deposition Modeling) methods.
  • Accomplished applications including hooked ball-mitten toys and hairy objects, requiring precise path control to avoid defects and achieve intricate designs.
  • Conducted experiments in 4D printing, modeling deformable planar objects in Fusion360 that transform into stereoscopic shapes when heated.
April 2022 - April 2023
Over-the-Air-Computation Based Federated Learning Model Establishment & Simulation
Zhejiang University (Advisor-- Prof. Howard Yang)
Edge Computing Federated Learning OFDM Over-the-air Computing Simulink Machine Learning Neural Networks
  • Explored an innovative approach to utilizing private data from distributed databases to train shared models, ensuring user privacy while making use of the data.
  • Set up an over-the-air-computation-based communication model in Simulink which could transmit and receive massive data gradients between federated users effectively.
  • Combined Machine Learning models like Linear Regression and Deep Learning models like neural networks in MATLAB codes with communication models in Simulink to implement effective edge-computing models.
  • Improved the model to adapt to the Large-scale applications by exploiting and modifying existing OFDM Communication Systems.

Honors and Awards

2025 Summer
  • Outstanding Graduate of Zhejiang University
2023 Fall
  • Dean's List (Top 20%) - UIUC
2021-2022, 2022-2023, 2023-2024 Academic Years
  • Scholarship from Zhejiang University (Top 8%)
  • Academic Excellence Award - Zhejiang University
2022 Summer
  • Recognition for Outstanding Performance, Undergraduate Summer Research Program
  • Recognition for Academic Poster Exhibition, Undergraduate Summer Research Program
2021-2022 Academic Year
  • Outstanding Performance in Social Work Award - Zhejiang University

Research Interests

  • Brain-Computer Interfaces (BCI) & Neural Engineering
    • Neural Signal Processing & Decoding: Advanced feature extraction and sequence modeling (LSTM/GRU/Transformer) for real-time speech/motor decoding from EEG/iEEG.
    • Multimodal Sensor Fusion: Integrating heterogeneous data sources (e.g., bio-signals, eye-tracking) to enhance BCI robustness and information transfer rates.
    • Neuromorphic Computing & RRAM: Exploring compute-in-memory architectures and resistive memory (RRAM) for energy-efficient neural interfacing.
    • Closed-Loop BCI Systems: Designing low-latency, feedback-driven systems for therapeutic applications and neuro-rehabilitation.
  • Integrated Circuits & VLSI Systems
    • Mixed-Signal IC Design: High-precision Data Converters (Delta-Sigma ADCs), PLLs, and Clocking circuits for biomedical applications.
    • Hardware Accelerators for AI: ASIC implementation of machine learning algorithms, focusing on power-efficient inference logic in advanced FinFET nodes.
    • Design-Technology Co-Optimization (DTCO): Power-performance-area (PPA) trade-offs in nanometer CMOS processes (180nm to 16nm FinFET).
  • Artificial Intelligence & Efficient Computing
    • Privacy-Preserving Machine Learning: Federated Learning (e.g., pFedMe) algorithms addressing data heterogeneity and class imbalance in distributed healthcare networks.
    • Edge AI & TinyML: Optimizing deep learning models via quantization, pruning, and hardware-aware training for deployment on resource-constrained embedded devices.
    • Computer Vision & Pattern Recognition: Real-time object detection and image reconstruction optimized for GPU (CUDA) and parallel processing architectures.
    • Explainable AI (XAI): Enhancing model interpretability in critical biomedical decision-making systems.

Other Interests

  • Basketball 🏀
  • Snowboarding 🏂
  • Hip-hop 🎤