yfyfgao.fyi
A LITTLE ABOUT YIFEI Open to PhD / MPhil opportunities
3D VISION × GPU SYSTEMS × AI AGENTS

Yifei
Gao.高逸飞

Building the systems behind
intelligent 3D reconstruction.

Information Engineering undergraduate
at Zhejiang University.

Explore my research Read my CV
FIELD STUDY / 0013D GAUSSIANS
YX
Drag to rotate · procedural illustration2,400 points
+ BASED IN HANGZHOU, CHINASCROLL TO EXPLORE

01 RESEARCH

A few things
I’m working on.

Efficient 3D representations.
GPU compute over Wi-Fi.
Research agents with feedback loops.

3 projects
FIG. 01SYSTEMS
GPU SYSTEMS · HKUST SPARK LAB

Bringing GPU power closer.

Remote GPU acceleration & cross-platform graphics

Keep data and code local. Run demanding GPU workloads remotely over commodity Wi-Fi. I adapted four 3D Gaussian Splatting frameworks and worked on remote graphics for macOS applications.

Jul — Sep 2026
Project notes

UG Visiting Research Student (HKUST) & R&D (WiCi AI)
Supervised by Prof. Zili Meng

Remote 3DGS training

Kept datasets and training code local while accessing remote GPUs over Wi-Fi. Achieved 83.28% of native training-loop throughput with gsplat, and 82.27% loop / 78.11% end-to-end efficiency with Original 3DGS. The remote setup delivered a 3.26× speedup over a local RTX 4060 with no quality degradation in the reported comparison.

Metal-to-Vulkan graphics

Developed a compatibility layer with remote rendering and video streaming. GravityMark 1.89 at 1080p, one million particles, and Temporal AA averaged 45.8 application FPS, versus 7.7 FPS on the MacBook Neo’s native Metal path: a 5.95× increase.

DaVinci Resolve workloads

Translated selected macOS Metal workloads to a remote RTX 5090 Laptop GPU. Exporting 500 frames of Standard Candle UHD: Rocket Science (66 Blur) at 1080p / 25 fps / AV1 took 6m 08s, versus 13m 33s natively on a MacBook Pro M5 Max: 2.21× throughput and 54.7% less export time.

3.26×3DGS speedup vs. a local RTX 4060
FIG. 02AGENTS
AI AGENTS · INDEPENDENT RESEARCH

Research, with a feedback loop.

research-harness · a multi-loop agent framework

A framework for long-horizon research that turns self-review into targeted revisions and execution failures into workflow repair. Task-specific skills stay modular; orchestration and recovery stay reusable.

Apr 2026 — PresentView code
Project notes

Independent Project
Supervised by Prof. Yiyi Liao

Inner loop: improve the work

Converts self-review into targeted revisions during research tasks.

Outer loop: repair the workflow

Uses interface contracts and execution tracebacks to guide workflow repair across a longer research process.

Pluggable capabilities

Separates task-specific skills from orchestration and recovery logic through a modular architecture.

2 loopsself-review inside, workflow repair outside
FIG. 03VISION
3D VISION · SRTP PROJECT LEAD

Reconstruct what matters.

Segmentation-aware 3D Gaussian Splatting

Use segmentation to focus reconstruction on the foreground. Removing irrelevant background lets the Gaussian budget work where it matters, with a diffusion workflow to augment sparse input views.

Mar 2025 — May 2026
Project notes

Student Research Training Program · Project Lead
Supervised by Prof. Jixing Cao

Foreground-focused reconstruction

Preprocessed training views with a segmentation model so Gaussians were allocated over the foreground. On a consumer RTX 4060, training time fell from 142 to 61 minutes (56.7%), L1 loss fell by 75.2%, and PSNR improved by 0.57 dB in the reported experiment.

Sparse-view augmentation

Built a hybrid 3DGS–diffusion workflow to augment 8–12 input images before training, evaluating reconstruction across varying structural complexity.

−56.7%training time in the RTX 4060 experiment

Performance figures refer to the experiments described in each project’s notes.

02 ABOUT ME

Curious about
the whole stack.

I’m Yifei Gao (高逸飞), an Information Engineering undergraduate at Zhejiang University.

My research connects 3D representations with the systems that make them practical. I’ve worked on segmentation-aware Gaussian Splatting, remote GPU acceleration, and agents for long-horizon research.

I want to advance efficient, intelligent 3D reconstruction through better algorithms, systems optimization, and autonomous experimentation.

The longer version, in my CV
EDUCATION

Zhejiang University

B.Eng. in Information Engineering

2023 — 2027Expected graduation · Jun 20273.71GPA / 4.0
TOOLS I THINK WITH

Languages

C / C++ · Python · MATLAB · Verilog · LaTeX · RISC-V Assembly

Vision & ML

3D Gaussian Splatting · OpenCV · ComfyUI

Hardware

Vivado / ModelSim · STM32 HAL · Altium PCB Design

Tools

Git · Linux · MATLAB Signal Processing Toolbox

03 BEYOND THE MAIN THREAD

Built to understand.

A few course projects, from processor
pipelines to underwater acoustics.

01OPTIMIZATION

PCB placement optimization

Feb — Jun 2026 · COURSE PROJECT

Greedy initialization, force-directed refinement, and two-stage simulated annealing with incremental wirelength evaluation and constraint-aware moves.

31.6%–83.1% less HPWL across five supplied benchmarks, with zero recorded spacing violations. Largest case: 115 components and 199 nets.

02COMPUTER ARCHITECTURE

A five-stage RISC-V CPU

Sep — Dec 2025 · COURSE PROJECT

Designed an RV32I pipelined CPU in Verilog HDL, with forwarding and stalling logic for data and control hazards.

Verified ISA compliance and hardware stability using testbenches in Vivado and ModelSim.

03SIGNAL PROCESSING

Underwater acoustic DOA

Sep — Dec 2025 · COURSE PROJECT

Implemented MUSIC and ESPRIT in MATLAB for high-resolution direction-of-arrival estimation from underwater acoustic arrays.

Included time-domain spectrum preprocessing for array signals.

04EMBEDDED SYSTEMS

Solar lighting controller

Jun — Jul 2025 · COURSE PROJECT

Built an STM32 controller with PWM charging, constant-voltage / constant-current management, and a 300 mA LED driver.

OLED battery display, keypad modes, timer-interrupt scheduling, and ADC-based closed-loop sensing using STM32 HAL.

RECOGNITION / 2025

National Encouragement Scholarship

Ministry of Education of China

Second-Class Scholarship

Zhejiang University

Real Estate Fund Scholarship

DUOYUAN GROUP & Dr. Ye Hongwei

04 WHAT’S NEXT

Let’s talk
research.

I’m seeking a PhD or MPhil position in efficient, intelligent 3D reconstruction. If our research interests align, I’d love to connect.

GitHub Curriculum vitae