My name is Li Duan, a quantitative developer with a strong background in software engineering and a deep interest in financial markets. I specialize in building high-performance systems for quantitative research and trading, with experience in low-latency infrastructure, algorithmic strategy implementation, and data-driven analytics. My journey spans across areas such as time-series modeling, high-frequency data processing, and GPU-accelerated computations. I’m passionate about writing clean, efficient code and continuously exploring new technologies in finance, mathematics, and machine learning. Let’s build innovative solutions for tomorrow’s markets together!
Download full resumeI'm passionate about developing robust, efficient and reliable systems while writing clean and well-structured code.
I employ a variety of tools to accomplish my objectives efficiently.
- Required courses include Computer System Architecture, Data Structures and Algorithms, Enterprise System Architecture, Software Engineering on Application Architecture and Artificial Intelligence. Advanced courses cover topics such as Big Data Systems in Data Science, Database Modeling and Programming, Real-Time Stream Processing, Financial Technology Solutions, Information Visualization and Digital Engagement.
- National patent: Dental veneer model generation and thickness measurement method based on oral scanning point Cloud.
- The software-related courses include C++ Programming, Data Structures and Algorithms, Data Statistical Analysis, Network Information Programming, Data Mining, Microcomputer Systems and Interfaces, Embedded Systems and Course Design, and Comprehensive Design of Digital Image Processing Systems.
- The hardware-related courses include Fundamentals of Circuits, Digital and Logic Design, Basics of Electronic Circuits, Principles of Automatic Control, Automation Components, Power Electronics Technology, Automatic Detection Technology, and Motion Control.
- Designed and developed a high-performance cryptocurrency trading gateway system that seamlessly integrates REST/WebSocket APIs for 5 major exchanges (Binance, Bybit, OKX, etc.). Leveraged C++23 features, multithreading, and async programming to ensure scalability and stability. Successfully deployed and maintained zero-downtime in production.
- Led development of the offline market data processing module and optimized real-time streaming pipelines. Refactored core components to enhance data processing efficiency and improve code structure. All changes passed 100% regression testing, ensuring 24/7 market data delivery with 99.99% stability.
- Responsible for the architecture development and upgrade of the Tencent Games Overseas Account Distribution System. Conducted an in-depth investigation into the advantages and disadvantages of various distributed frameworks (Nacos, Consul, Zookeeper, Eureka, Etcd, etc.) and strategically selected Consul to implement unified service registration, discovery, health check and service mesh(sidecar proxy) across both ECS and Kubernetes environments. Utilized Consul KV for storage of service configuration, routing, and weight information.
- Redesigned Tencent's self-developed load balancing agent and configuration agent, focusing on high-performance, low-latency operations, to adapt to the account system's configuration center and overall architecture. Implemented a multithreaded mechanism real-time updates of configuration and routing data stored in Consul to the configuration center agent, shared memory, and subsequently loading it into the load balancing agent. Led solution evaluation, development, self-testing, and benchmark testing. Successfully deployed across both ECS and Kubernetes environments.
- Designed a protection strategy for sensitive interfaces, implemented mTLS(HTTPS mutual authentication) on load balancers in TCC, AWS, GCP, and Azure, replacing manual IP whitelisting. Tested the functionalities of mTLS, certificate revocation, periodic certificate rotation, and multi-certificate chains. Additionally, developed a backup strategy by leveraging Cloud LB to forward Layer 4 requests to an Nginx server, where mTLS functionalities were configured on Nginx to enable mutual authentication.
- Developed and implemented a customized Umbraco Content Management System (CMS) tailored for financial business management, featuring MySQL integration and advanced access control. The solution provided seamless content management capabilities for other modules within the platform and was successfully deployed in the latest release of the financial management system, now fully operational.
- Developed a Linux kernel-space device management driver and a user-space device simulator (C/C++), using interrupts to verify the behaviors and states of the device simulator. Led solution formulation, architecture design, module development, and integration testing. Ensured stable operation of the driver on next-generation server CPUs (GNR).
- Optimized data-parallel algorithms for Intel's high performance GPU architecture using oneAPI, rewrote critical sections of the parallel algorithms with SYCL (C++) to ensure low-latency, high-throughput execution, achieving a 5-10x reduction in runtime compared to previous cross-platform parallel algorithms on PVC GPUs. Analyzed assembly code of parallel algorithms to propose impactful enhancements for the next-generation PVC GPU compiler, driving further optimization and efficiency.
- Responsible for functional testing of BIOS architecture and RAS modules. Designed test cases for BIOS architecture upgrades targeting next-next-generation CPUs (DMR), using the Simics simulation platform to cover both software simulation and hardware integration environments.
- Contributed to the development of the Cloud Xiaowei Open Platform and voice assistant microservice modules, utilizing the trpc-go framework for service communication. Developed protocol integrations for speech recognition, semantic parsing, and speech synthesis modules. Additionally, managed Redis data migration and cluster deployment.
- Refactored the sports open capability module from C++ to Go services. Spearheaded the project for visually impaired programmers, coordinating with company representatives to provide Python cloud development tutorials on the open platform, while gathering feedback on accessibility support for the Cloud Xiaowei Open Platform.
- Developed a lightweight JavaScript compilation engine backend for the Taobao app, focusing on optimizing performance. Led the development of instruction bytecode and JIT compilation modules, driving improvements in execution speed and resource efficiency, achieving a 10% increase in execution speed and enhancing resource efficiency by 7%.
- Developed the file clean-up module for online bypass repair using Java, overseeing the processing of handheld local and platform-delivered files in real-time, aligning with dynamic business push data to achieve seamless hot repairs, resulting in a 15% improvement in repair response times.
Feel free to explore my GitHub
OfferAI
- A job-seeking assistant platform allowing users to register and utilize AI-based features such as resume enhancement, job recommendations, interview preparation, cover letter generation, and job applications. It also provides comprehensive job application tracking, job search, and matching functionalities.
Rate Corner
- A course evaluation platform allowing students to register, provide course evaluations, view and filter other course reviews, and manage users' own assessments. Widely used by NUS students, the platform has garnered thousands of favourites on social media promotional posts.
Tianjiang Food Delivery Platform
- The project is a customized software product specifically designed for catering enterprises such as restaurants and hotels. It comprises a system management platform and a mini-program application on the client side.
Segmentation technology of mandibular neural tube in CBCT cross-sectional image
- Utilized segmentation algorithms based on 2D-UNet and 3D-UNet neural networks. Led the solution design, literature review, implementation of deep learning algorithms, and software integration development. Developed using Python and C++.
- Achieved a segmentation accuracy of 99.7% and a 3D model overlap rate of 85%, surpassing traditional algorithms by over 10%. The software has been deployed and utilized at Nanjing Stomatology Hospital.
CBCT Image Segmentation and 3D Reconstruction Based on Deep Learning
- Collaborated with Nanjing Stomatological Hospital for software for maxillary sinus segmentation and 3D reconstruction. Applied deep learning with neural networks such as Mask R-CNN, U-Net, and U-Net++ and algorithms including Marching Cube, Laplace to enhance precision and efficiency.
- Responsible for research and implementation of image segmentation neural networks and reconstruction algorithms. Accomplished an outstanding accuracy of 99.9% for the segmentation and an impressive 93% overlapping rate for 3D models, surpassing traditional algorithms by over 15%.