Hi, I'm Khang

In the early years of my career (2022), I focused on mobile development, building high-performance and smooth Android and iOS applications with beautiful user interfaces. Then, OpenAI popularized the LLM era, simplifying my workflow and giving me more time to research and expand my tech stack, leading me to delve into Web, Backend, and AI.

Between 2024 and 2025, I participated in image processing projects such as Face Aging, Image to Sketch, Image Brightness Enhancement, RGB Image to Depth, reconstruction tasks, image generation, Pic2Pic models (like image to anime), Stable Diffusion-based models, object detection, and segmentation models, gaining extensive experience in Computer Vision.

Amidst the rapid growth of LLMs, I read a paper by AI pioneer Yann LeCun addressing the limitations of current LLM architectures and proposing the concept of World Models—AI systems capable of acquiring "Common Sense" (fundamental real-world understanding). These models learn by watching unlabeled videos and images, mimicking human learning via Computer Vision, Self-Supervised Learning, and future prediction. Yann LeCun is also actively bootstrapping a research startup focused on this topic.

This inspired me to pursue a Master's degree in Computer Science at UTH University, focusing on Computer Vision research outside of work. Currently, my primary role is mobile app development and training visual AI models. Besides heavy server-side models, I specialize in training and optimizing lightweight models to run directly on client devices, saving server resources and optimizing operational costs.