• Automated the performance analysis of CPU performance in programs involving computer vision and shortened the analysis time of CPU performance significantly from weeks to less than 5 minutes.• Enhanced the instruction level simulator in C to capture trace data for the analysis of CPU performance across L1, L2, and L3 cache layers, driving the identification of bottlenecks and improvement of performance.• Designed a script in Python that produced user-customizable graphs visualizing the CPU performance, which received positive feedback from multiple senior members of the team.• Generated XML files containing all the relevant information from the trace output analysis for further simulation.

• Contributed to Autodesk Shape Manager (ASM), the geometric modeling kernel behind major Autodesk products (Fusion 360, AutoCAD, Inventor) with millions of subscribers, using C++ (OOP-based).• Tested a new feature of ASM (asymmetric blending of edge), identified and documented bugs, and presented findings to the international team, facilitating the continued development of the feature.• Researched methods to detect unusual changes in code performance more accurately and designed custom algorithms with x2 sensitivity; discussed findings and shared insights with senior colleagues.• Reduced processing time of surface creation by 40% by implementing multi-threading.• Created new wrapper classes which reduced memory usage and decreased processing speed by 60%.• Wrote 80+ new unit tests using Scheme and C++, each covering different scenarios, to ensure robust and error-free code.

• Cooperated with other team members to build and maintain a humanoid robot (InMoov).• Implemented OpenCV and MediaPipe for pose and face detection.• Manipulated data of 30+ points for the calculation of joint angles and movement with Python.

• Designed an application in Python as a team of 3 for a start-up for rural and suburban car-sharing.• Submitted up to 1000 API calls to Google Maps API (Directions and Distance Matrix) to obtain relevant data.• Devised and tested an algorithm that matches drivers and riders and produced a prototype that matched 200+ riders to drivers based on pre-set preferences and geographical locations.

• Won 1st Place in Hong Kong Regional Contest in ABU Robocon 2021• Led a team of 2 to design and build a motor-based shooter with up to 90% accuracy within 4 m• Programmed in Arduino to control up to 10 pneumatic cylinders and motors through PID control