Maintain, debug, and develop SLAM algorithms for Virtual Reality (VR) headsets.- Maintain the "Double Tap" detection algorithm for VR headsets (a) Implement testing processes and associated GitLab-CI continuous integration scripts (b) Collect data and construct a database for GitLab-CI testing in collaboration with colleagues from the software testing team (c) Improve algorithm precision and recall rates by adjusting parameters based on GitLab-CI testing results- Maintain and reimplement the “high frequency” module in the SLAM algorithm of the VR headset (a) Master and understand the principles and codes of the “high frequency” module (b) Debug and fix issues fed back from software testing team colleagues and customers (c) Reimplement the “high frequency” module to streamline the code and improve code readability, facilitating subsequent maintenance and development (d) Reduce the output jitter of high-frequency poses via the Kalman filter- Implement a Visual-Inertial Odometry via the marginalization strategy and the “SLAM points” method to improve the user experience of VR headsets (a) Master and understand the principles and code of existing Visual-Inertial Odometry (b) Identify input and output data required by different modules and confirm data interfaces with collaborative colleagues (c) Derivate and implement mathematical formulas (d) Test and validate the implemented code objectively and subjectively

Worked on cutting-edge research on SLAM and event-based vision algorithms, and provide technical support for product developments. Implemented a 6-DoF visual odometry with a high tracking speed (~300 FPS) and improved an event-based optical flow program to speed up the computation by 5 times- Studied and implemented the ICE-BA algorithm, a state-of-the-art incremental Bundle Adjustment method to speed up the back-end computation of Visual/Visual-Inertial Odometry- Participated in a VR headset “See-through” project and applied image stitching algorithms- Implemented a 6-DoF visual odometry based on the inverse LK optical flow method, achieving a tracking speed of up to 300 FPS- Participated in an AR optical see-through calibration project and wrote the optimizer- Implemented and improved event-based vision algorithms: (a) Implemented an event-based visual odometry using the Event-based Multi-View Stereo (b) Supervised an internship to implement ELiSeD, an event-based line segment detection algorithm (c) Improved an event-based optical flow program by replacing the Ceres automatic differentiation with explicit closed-form expressions and using Intel AVX instructions to speed up the computation by more than 5 times

Accomplished the Ph.D. project “Resilience Analysis of Nuclear Fuel Cycle Scenarios” and developed a novel framework for the design of complex nuclear fuel cycle scenario systems:- Developed an innovative Bayesian optimization algorithm to rapidly estimate the region of feasible solutions for a complex system with multiple constraints- Implemented the developed Bayesian optimization algorithm on the URANIE statistical analysis platform using C++, and performed its multi-CPU parallel optimization- Developed a nuclear reactor fleet evolution algorithm based on C++, which can characterize and determine the complex evolutionary history of a nuclear reactor fleet with few input parameters- Constructed an analysis paradigm for nuclear fuel cycle scenarios to analyze the resilience of different nuclear fuel cycle strategies under uncertainties- Wrote Python scripts to analyze data and visualize the study results- Wrote technical reports and the Ph.D. manuscript in English. Successfully defended thesis.

Designed reactor cores and core reloading patterns and provided on-site technical support for nuclear power plants

Developed tools to perform a large number of physics simulations, including communication of models, automatic execution of codes, and the neural network used as surrogate models for time-consuming physics simulation