Troy Gu

This AI system identifies the chicken eggs' gender before they hatch and are loaded into the incubator. I calibrated a hyperspectral image, analyzed data, designed and extracted features, and trained and optimized our AI algorithm and models. I also built deep learning models using the technology of 1D, 2D, and 3D convolutional neural networks (CNN) and various long short-term memory (LSTM) in the cloud environment. Finally, I reduced the space of features and fused models.

This real-time portable device evaluates the quality, including the marbling score and color of pork chops, through the RGB camera. I headed a five-person team to build this device from scratch, including selecting the hardware platform and developing a lighting system and an application program. Also, I segmented the pork chops' region of interest (ROI), trained and optimized models, ported coded and optimized algorithms, and transferred codes from MATLAB to Python and then to C++ to reduce the overall time.

Our team built a PIC-E card, which was used to protect input and output data on the host machine. We used the PCI-E interface, DSP, FPGA, and secure chips.

My duties involved:
• Participating in a complete product research and development (R&D) cycle, from a concept to the final production, including the architecture design and key chips selection.
• Developing a device driver, embedded software, a PCI-E card driver for Linux and Windows platforms, the program running on the card in a bare-mental environment, a boot loader program, an application program, and an API.
• Completing a card fault diagnosis program independently to control the production quality. The program can diagnose and accurately position most faults at the hardware level, such as shorting pins and pin non-welding.
• Building card testing tools.

As an embedded software engineer working on this project, I developed a Linux driver and test tools for the audio module of ARM system-on-a-chip (SoC) to verify all data paths of audio register functions, including memory to speaker and microphone to memory.

While working as a student researcher at Lenovo, I oversaw the Android platform multi-language module development, search module for file management software, and file compression module coding running on Lenovo LePad, a commercial mobile product.