Resume

Software Engineer

Aurora Innovation

Built high-performance and secure software systems and improved the framework which AV Motion Planning Algorithms integrate.

• Implemented robust interfaces in C++ to represent autonomous vehicle environment and ego status by integrating online map and sensor data, and serialized information into protobuf format for downstream planning algorithms and visualization.
• Optimized trajectory ranking latency by extracting strategy-free boundary data, reducing tensor size in CUDA to eliminate redundant boundary violence cost computation across multiple strategies.

Software Engineer

Hughes Network System

Designed and implemented software for the low-latency LEO satellite user terminal system.

• Enhanced FPGA interfaces in C++ to interact with 10+ PCIe devices on the satellite user terminal, a real-time embedded system, conducted unit-test with CTest, resulting in a substantial increase in code coverage from 60% to 90%.
• Elevated hardware communication and monitoring concurrency by optimizing synchronization and memory mechanisms to achieve a 10ms responsiveness target for mobile features, a reduction from 100ms in stationary counterparts.
• Created 20+ REST APIs using ExpressCpp to access over 130 device statuses and passed tasks through ZeroMQ.

Software Engineer Intern

WeRide.ai

Worked on the Perception Infrastructure of the autonomous driving system.

• Integrated safety module of the perception infrastructure with C++ and Python to rightly detect and handle over 40 fault scenarios, reduce 17% of false alarms and record informatic fault history efficiently with limited hardware resources.
• Enabled 3 fault injections to trigger sensor input, intra-module, and system level failures on simulation machines, analyzed the error reports and system metrics using Grafana to validate the injection results.
• Developed diagnostic program monitoring GPU with NVML and recording delay accounting to triage 3-level failures.
• Extended Failure Mode and Effects Analysis (FMEA) scenarios from 14 to over 200 for the autonomous driving system.

Software Engineer Intern

Intel Asia-Pacific R&D

Tested and improved ML framework inference performance on Intel platform.

• Built out-of-box automation test scripts in Bash and Python to evaluate the installation and operation efficiency of six popular machine learning frameworks and products of Intel and Nvidia.
• Utilized Dockerfiles to build containers with suitable environments to facilitate the efficiency test as well as for 30+ deep learning workload running with 6 machine learning frameworks.
• Enabled the throughput of Keras workloads that under tf1.x in FP32 improve 12%-25% under intel-tf2.4 in BF16.

Backend Software Engineer Intern

Tencent

Developed backend service for WeChat.

• Created gRPC services in C++ to clean overdue records from large tables which kept inserting new records.
• Adopted crontab command and Bash scripts to simulate the million-class process of inserting records.
• Wrote multi-threading Bash scripts that called the cleaner service in parallel, saving an average of 70% of time.
• Reduce the lock impact of mass deletions by 40% by optimizing the queries and selecting InnoDB engine for MySQL.

Master of Science

Washington University in St. Louis

Computer Science

Bachelor of Science

South China University of Technology

Computer Science

Django REST API Project

A Django based backend REST API service.

• Utilized Django REST Framework to develop REST APIs for DoorDash-like ordering web service following the MVC pattern.
• Improved API security by applying token-based authentication, authorization and throttling for 4 user groups with Djoser.
• Designed relational data model stored in the SQLite3 database and implemented serializers for each data table.

Web3 Chat Room

A social platform for NFTs collectors based on Firebase service.

• Devised UX & UI of a chat room platform on Figma for NFT collectors to communicate and display their NFTs.
• Refactored and cleaned the front-end structures and features in React.js and Tailwind CSS based on 3 rounds of user tests.
• Launched real-time online users and activity features for 70+ users in Node.js and Next.js with the NoSQL database Firestore.