Kelan
Liu

Software Engineer

• ·Architected and implemented a serverless translation management system supporting localization across multiple product front ends; delivered a department-wide presentation on problem context, architecture, and system design (Python, TypeScript, AWS Lambda, S3)
• ·Collaborated cross-functionally across teams to ship features end-to-end, contributing directly to adjacent team codebases
• ·Developed backend APIs and event-driven automation workflows using Node.js, TypeScript, and Python, deployed on Amazon ECS and AWS Lambda
• ·Optimized critical SQL queries, reducing database query costs by up to 68% (MySQL, Redis)
• ·Built a resilient request library with retry logic and exponential backoff, reducing socket hangups and server-side errors
• ·Identified and patched a privilege-escalation vulnerability caused by misconfigured AWS SES IAM permissions
• ·Built a multi-agent AI debugging tool using Claude Code with a log-tailing watcher agent, interactive code-fix session, and automated QA agent for curl-based validation (Python, Bash)

Software Engineer, Machine Learning, Research Intern

• ·Contributed to a real-time autonomous driving simulation platform through kinematics modeling and cloud visualization integration (C++, Python)
• ·Partnered with Simulation and UI teams to refine scenario realism and optimize user interaction workflows, improving fidelity across kinematics modeling, visualization, and driving UI
• ·Built an operator-facing driving UI and a data collection pipeline for simulation workflows (Angular, RxJS, TypeScript, Protobuf)
• ·Optimized end-to-end latency to under 200ms by implementing D3.js-based visualization tools for monitoring rendering and mapping plugin performance
• ·Significantly improved intern permission pitfalls by implementing intern-friendly scripts for existing software infrastructure

Software Engineer Intern

• ·Ported and optimized video-processing and ML algorithms for embedded GPU systems using CUDA, Docker, and Python
• ·Implemented and optimized real-time motion detection algorithms supporting a YOLO-based license plate detection pipeline for the Jetson Nano and Xavier (OpenCV)
• ·Reduced object detection and classification training runtime by 4 minutes on a large training dataset of thousands of vehicles