林煜轩 Yuxuan Lin

yl6061@columbia.edu  |  +1 (646) 705-7015   

Hi! I am a M.S. in Computer Engineering at Columbia University, expecting to graduate in December 2026.
Before that, I received my B.S. degree in Computer Engineering from University of Illinois Urbana-Champaign and a B.Eng. degree in Electronic & Computer Engineering from Zhejiang University.

LinkedIn  /  Github  /  Blog

Research Experiences

My research explores generative models for medical imaging and 3D content creation.

• Formulated PET(Positron Emission Tomography) reconstruction as a Bayesian inference problem, using conditional normalizing flows (RealNVP, Glow) to model the posterior distribution of tracer activity.
• Implemented a Deep Probabilistic Imaging (DPI) pipeline that outputs both posterior mean reconstructions and uncertainty estimates.
• Scaled training with multi-GPU parallelization (3.7x speedup) while keeping image quality stable (<1% drop).

• Built a GenAI powered Blender plugin with Python, supporting 3D prototype management, segmentation, and Gaussian ↔ Mesh conversion.
• Enabled text-to-3D model generation using Transformer-based Gaussian Splatting and mesh rendering pipelines.
• Deployed a Flask-based backend to track user interactions and support 3D generation services.

Professional Experiences

I had fun doing internships in software development.

• Developed control system modules for commercial cleaning robots on FreeRTOS and Linux OS.
• Designed and implemented sensor logic for GD32 microcontroller using the Keil environment, ensuring precise control and real-time operation.

• Led bi-weekly discussion sessions to strengthen students' grasp of core discrete mathematics concepts.
• Contributed to the design of course materials, including homework and exam samples, ensuring alignment with learning objectives.
• Graded assignments and exams with accuracy and provided constructive feedback; supported students through office hours.

Selected Projects

I am familiar with C/C++ and Linux.
Also, I have experience in CUDA, Golang (Gin, GORM), Python (PyTorch, Flask), SQL/NoSQL, x86 Assembly, and UE5.

• Implemented and optimized the forward-pass of a convolutional layer of the modified LeNet-5 using CUDA
• Used Streams to overlap computation with data transfer
• Used Tensor Cores and Shared Memory Tiling to speed up matrix multiplication with an advanced GEMM kernel
• More optimizations included unrolling, constant memory, Fixed point (FP16) arithmetic ...

• Developed a Linux-like operating system kernel from scratch using C and x86 assembly language
• Supported interrupts, system calls, exceptions, managed by 8259 PIC; Implemented kernel and user state switching logic, using TSS - task state segment to support IRET.
• completed the virtual memory (using page directory/table), file system (two-layer mapping), terminals for display and so on; Currently supported devices include keyboards, RTC, and PIT interrupts
• Multi-process scheduling with scheduler and multi-terminal switching (buffer, cursor positions maintained by PCB - process control block)
• Led a team of four and used common development tools such as Git for team version control and GDB for debugging

• Developed an adventure puzzle-solving game demo in a team of 5, themed after Infinity Blade. Utilizing Unreal Engine (UE) with its Blueprints visual scripting system.
• Implemented core gameplay mechanics including health system, collectible items, and created AI-controlled pursuer enemies, mortar enemies, player-enemy collisions for interaction.
• Built four integrated levels—Lava Parkour, Laser Puzzle, Riddle Maze, and Traffic Jam to deliver diverse and immersive gameplay experiences.

Relevant courses:

Source code from Jon Barron.