Hsiang-Jui Lin (Jerry Lin) | 林祥瑞
Email: jerry73204@gmail.com | d10922013@csie.ntu.edu.tw
GitHub: github.com/jerry73204 | LinkedIn: linkedin.com/in/jerryhj-lin
PhD candidate at National Taiwan University specializing in autonomous vehicle systems, multi-modal sensor fusion, and V2X communication. Systems programmer with expertise in Rust, C/C++, and distributed consensus algorithms. 5+ years of industry experience in R&D roles at Microsoft and ADLINK Technology, with a focus on scalable backend systems and real-time networking.
Education
- PhD Program in Computer Science and Information Engineering, National Taiwan University (2021–Present): Advisor: Professor Chi-Sheng Shih | Research: Autonomous vehicles, sensor fusion, V2X communication
- Master of Science in Computer Science and Information Engineering, National Taiwan University (2021)
- Bachelor of Science in Computer Science and Information Engineering, National Taiwan University (2016): Thesis: Solving Graph Domination Problems on Distance-Hereditary Graphs Using Split Decomposition
Research Projects
- AutoSDV (Autoware Software-Defined Vehicle) (2023–Present): Main contributor to AutoSDV, an open-source autonomous driving platform. Designed and implemented the complete vehicle architecture and software stack based on Autoware. Developed hardware integration for practical sensors, enabling affordable self-driving research platforms for educational institutions.
- Multi-Modal Autonomous Vehicle Communication Control (2023–2024): Led development of innovative communication protocols for heterogeneous autonomous vehicle platforms in collaboration with KingwayTek (勤崴國際).
- Real-Time Perception and Collaborative Driving (2021–2023): Developed V2V cooperative decision-making algorithms and collaborative perception systems with National Yang Ming Chiao Tung University.
- Cross-Domain Time-Sensitive Networking (2022–2023): Researched deterministic networking protocols using time-sensitive networking (TSN) technologies for vehicular applications with ADLINK Technology (凌華科技).
- Smart Cockpit Interactive Systems (2023–2025): Leading research on gesture-based interaction and spatial audio systems for next-generation vehicle interiors with AU Optronics Corporation (友達光電).
Publications & Presentations
Publications
- Unsupervised Training Framework for 3D Point Cloud Object Detection Model (2025): Developing self-supervised learning approaches for 3D object detection in autonomous driving applications.
- Self-Supervised Multi-LiDAR Object View Generation Using Single LiDAR (2023): Proposed novel method for generating multi-view LiDAR representations from single sensor input, improving perception robustness.
- A Performance Study on the Throughput and Latency of Zenoh, MQTT, Kafka, and DDS (2023): Comprehensive evaluation of pub/sub frameworks demonstrating significant throughput improvements in specific scenarios. (1 citation)
- Scalable and Bounded-time Decisions on Edge Device Networks using Eclipse Zenoh (RTSCA 2022): Developed consensus algorithm achieving deterministic decision-making within 100ms bounds. (4 citations)
- Prediction of Human Intention in Vehicles, Pedestrians and Bicyclists Interactions (IEEE ITSC 2021): Developed deep learning models for predicting road user behavior, improving autonomous vehicle safety. (1 citation)
- Master's Thesis: Occlusion-Resistant Tracking Based on Idempotent Adjunction Relationships (2021): Achieved 0.3m average displacement error in vehicle tracking under 70% occlusion conditions.
- Undergraduate Thesis: Graph Domination Problems on Distance-Hereditary Graphs (2016): Discovered O(n²) algorithm improving upon previous O(n³) solutions.
- 機器人程式設計與實作:使用 Java (Robotic Programming Design and Implementation Using Java) (2013): Published technical book on LEGO NXT robotics programming. ISBN: 9789862768228
Presentations
- ROS2 ♥ Rust (COSCUP 2024): Presented integration techniques for using Rust with ROS2 (Robot Operating System 2), showcasing type-safe robotics programming and performance benefits. Session details
- Highly Performant Dataflow in Stream-Oriented Programming with Rust (COSCUP 2023): Demonstrated significant performance improvements using async stream processing. Session details
Professional Experience
- Research & Development Assistant, ADLINK Technology Inc. (2021–2024): Developed distributed consensus algorithms for Eclipse Zenoh middleware, achieving sub-millisecond latency for edge computing applications.
- Research & Development Assistant, Microsoft Taiwan (2018–2019): Optimized Bing's reverse geocoding backend services, improving query response time and system scalability.
- Software Developer, CAVEDU Education (CAVEDU 教育團隊) (2010–2016): Led AppInventor curriculum development. Authored technical documentation and conducted workshops on robotics programming.
Technical Skills
Programming Languages: Rust, C/C++, Python, Java, CUDA, OpenCL
Frameworks & Tools: Autoware, ROS/ROS2, PyTorch, TensorFlow, Eclipse Zenoh, DDS, Docker, Git
Specializations: Distributed Systems, Parallel Computing, Computer Vision, V2X Communication, Real-time Systems
Open-Source Projects
- AutoSDV (2022–Present): Main contributor to the Autoware Software-Defined Vehicle platform. Provides a complete autonomous driving solution from hardware specifications to software implementation, democratizing self-driving research for educational institutions.
- ddshark (2023): Real-time RTPS protocol analyzer similar to htop, supporting multiple DDS implementations.
- Additional projects: github.com/jerry73204 — Including carla-rust, and other repositories.