Hi, my name is Eric Qiao
I'm an AI Solutions Engineer.

I built Mr.Gingerpaw in my spare time to solve a simple but annoying problem: keeping track of my kitchen ingredients. As the main cook, I often forget what I already have, which leads to extra shopping trips and food waste.

Mr.Gingerpaw is a full-stack project I put together end-to-end: a FastAPI + PostgreSQL backend, a React frontend, and automated CI/CD on Azure using GitHub Actions. It was a great way to learn new tools and practices quickly.

It’s still a prototype, but I’m already planning to add AI features—like NLP for smart shopping lists and image recognition so I don’t have to enter every item by hand.

You may have heard of Kinect or Realsense depth cameras, which capture images with depth information. In my PhD research, I worked on the core technology behind those sensors -- structured light systems.

Structured light is a classical technique for depth sensing, used in various applications, such as 3D scanning, robotics, and intelligent manufacturing. I took a novel approach by representing 3D shapes with neural signed distance fields (NSDF) and training a neural network to reconstruct those shapes from the captured images. We were the first to apply this framework to structured light systems, and the results were very promising.

I won't go much into the details here, but you can find the full paper (published in 3DV 2024) at the link below.

I developed a novel method for depth estimation in structured-light systems that works reliably in dynamic scenes. Unlike many deep-learning approaches, our system adapts its parameters on the fly—an essential capability when things are moving.

The core idea is to fully utilize the temporal information: motion isn't just a challenge, it's an advantage. We extract sparse flow trajectories between frames and use them to update the model parameters in real time. The result is high-accuracy, real-time depth-sensing system. (Published in IROS 2023 and RA-Letter 2022)

As a big fun of games, I often come up with ideas to improve the experience. That's why I made this mod for Monster Train, a deck-building roguelike. It adds a new clan with fresh mechanics and cards into the game.

It was not an easy task -- as I had no prior experience in game modding or C#/Unity. But by learning from the official wiki, diving into the game's source code, and turning to a really helpful community, I got it working. I'm quite proud of what I accomplished!