Deep learning research engineer with over 3 years of experience in computer graphics and geometry processing, applying cutting-edge deep learning techniques to solve real-world business problems. I thrive on delivering impactful products, having iteratively deployed and improved solutions based on user feedback in 3D asset generation and optimization. My expertise lies in seamlessly blending advanced technology with practical applications, driven not by the tools themselves but by the desire to create efficient, user-centered solutions that address core business needs.
-
Deep Learning Researcher, KRAFTON Inc., Seoul, South Korea (Nov. 2022 - Present)
- Working with diffusion models with ControlNets and LoRA layers for multi-modal PBR texture generation.
- Optimized a watertight mesh conversion algorithm to obtain target occupancy fields for mesh generation models, reducing processing time from hours to seconds per asset (cubic to quadratic time).
- Engineered large reconstruction models for single-image 3D reconstruction, facilitating features like the 3D Printer in inZOI.
- Implemented image-based texture mapping using ray tracing, drastically improving texture quality with less than one second, successfully deploying it to the 3D Printer in inZOI.
- Automated the standardization of large-scale 3D assets from online sources (e.g., Objaverse) and in-house datasets for model training, including processes like retopology, UV merging, and conversion to Principled BSDF.
-
Machine Learning Engineer, Spacewalk Inc., Seoul, South Korea (Jul. 2021 - Sep. 2022)
- Continuously deployed reinforcement learning agents (e.g., A3C, PPO) that complies with legal regulations while presenting architectural designs optimized for maximum profit.
- Designed a pairwise ranking learning pipeline to automatically shape reward functions, reducing tuning time by a factor of 200 and improving alignment with expert preferences from 55% to over 90% (F1 score).
-
MSc in Computer Science, KAIST, Daejean, South Korea (Sep. 2019 - Aug. 2021)
- GPA: 4.25/4.30
- Advisor: Prof. Sung-Eui Yoon
- Thesis: Deep Light Clustering for Denoising Monte Carlo Renderings
- Research area: physically-based rendering, deep learning, image denoising, clustering
- Coursework: advanced computer graphics, advanced deep learning, machine learning for 3D data, gpu programming, design and analysis of algorithm
-
BSc in Computer Science & Mathematical Science (double major), KAIST (2015 - 2019)
- GPA: 4.00/4.30
- Honors: summa cum laude, dean's list (top 3%)
- Coursework: computer graphics, operating systems, real/complex/numerical analysis, linear algebra, matrix computation, probability theory, statistics
- Email: [email protected]
- Linkedin: In-Young Cho
- Study Log: Notion
The image was created using ChatGPT-4 to illustrate a scene that matches the meaning of my name, which is benevolence (ไป) and prosperity (ๆฆฎ).