Shelf Object Detection

Developed a real-time Python program to automate shelf inventory monitoring, reducing manual effort and errors in store management.
Optimized image preprocessing using OpenCV and fine-tuned a YOLO pre-trained model for accurate detection in high-variance environments.
Achieved 90% detection accuracy and sub-100ms processing time per image, enabling scalable retail solutions.

#Shelf Monitoring #Python #YOLO #OpenCV #Deep Learning Code

Saliency Maps

Developed a Python-based tool to generate saliency maps, identifying influential pixels for image classification tasks in the Cats vs Dogs dataset.
Implemented and fine-tuned a convolutional neural network with TensorFlow and Keras, leveraging Class Activation Maps and gradient-based techniques for visualization.
Enhanced model interpretability by overlaying saliency maps on original images, enabling precise validation of feature focus.

#GradCAM #TensorFlow #Image Classification #Model Interpretability Code

Stylized Image Generator

Implemented a Neural Style Transfer pipeline using InceptionV3 model, selected layered to extract features, and optimized generated images to achieve stylized content.
Developed custom functions using TensorFlow for image preprocessing, calculating Gram matrices to capture style features independent of spatial structure, and ensuring the fidelity of source images.
Converged the losses using Adam optimizer with exponential decay learning rate to produce more appealing results within the same generating duration.

#Neural Style Transfer #Gram Matrices #TensorFlow #Generative AI Code

Little GO Game

Pruned search spaces by implementing Minimax algorithm with Alpha-Beta pruning for decision-making in the Go game.
Tuned weights dynamically for different phases of the game with evaluation function combining liberty, territorry, captures, and control.
Adopted a Q-learning approach to store historical data about games, enhancing AI performacne over time.

#Min-Max Algorithm #Alpha-Beta Pruning #Reinforcement Learning #Python

Experience

Software Developement Engieer Intern @ Amazon Web Services

Jun 2025 - Aug 2025

At Amazon AWS Lambda, I built an MCP-based troubleshooting tool integrated with Amazon Q Chat to automate Event Source Mapping (ESM) troubleshooting. I collaborated with the Stream Poller team to design scalable event-driven workflows using Kafka, Kinesis, and DynamoDB Streams.

Image Signal Processor Architect @ MediaTek

Oct 2016 - May 2024

At MediaTek, I served as an Image Signal Processor Architect, leading a team in evaluating performance for cockpit SoCs and collaborating on AI-friendly architectures for flagship mobile SoCs. I drove innovation through new ISP designs and automated data analysis workflows, enhancing efficiency and scalability.

Featured Projects: Dimensity 9000, Dimensity 9400

Image Signal Processor Engineer @ Genesys Logic

Oct 2013 - Aug 2016

At Genesys Logic, I worked as an Image Signal Processor (ISP) Engineer, focusing on streamlining verification processes for ISP designs. I optimized workflows and collaborated on enhancing the efficiency and reliability of hardware development.

Research Assistant @ CY Lab

Sep 2011 - Jul 2013

At CY Lab, I worked as a Research Assistant in Computer-Aided Design (CAD), focusing on algorithms for circuit optimization and verification. I devised an SAT-based algorithm and publised the findings in SOCC 2015.

Publication: Synthesis and Verification of Cyclic Combinational Circuits

Software Engineer Intern @ UMC

July 2011 - Aug 2011

At United Microelectronics Corporation (UMC), I worked as a Software Engineer Intern, where I focused on developing and optimizing software tools to support semiconductor manufacturing processes.

Portfolio