Jingtao (Tommy) Liu

Conducted Human-Computer Interaction (HCI) research and full-stack software development, focusing on advanced image processing, scalable research platform architecture, real-time collaboration tools, and robust remote device management systems.

• Designed and implemented advanced image processing algorithms, co-authoring a peer-reviewed paper published at an ACM conference.
• Iterated on and refined the full-stack architecture for a National Science Foundation (NSF)-funded research platform, significantly enhancing efficient on-site data collection and real-world deployment capabilities in educational settings.
• Developed and deployed a real-time WebRTC video calling application for remote collaboration, leveraging React, TypeScript, and WebRTC APIs; built custom signaling logic and media stream handling with a Python Flask backend on AWS (EC2, S3) for secure credential management and session control.
• Architected and delivered a cloud-based remote device management system utilizing MQTT, React, Node.js, and SQLite, providing real-time status monitoring, remote command execution, and multi-device coordination through a centralized dashboard for distributed hardware units.

Co-founded and served as CTO of an EdTech startup, leading cross-functional R&D teams and overseeing the development and deployment of multiple core web platforms and developer tools for K-12 STEM and AI education.

• Built and led a cross-functional R&D team encompassing frontend, backend, hardware, and mechanical engineering, implementing weekly code reviews, sprint syncs, and feature planning to ensure delivery velocity and code quality.
• Designed, prototyped, and successfully shipped core web platforms adopted by over 500 K-12 schools, including:
• Developed CocoBlockly, a browser-based visual programming IDE for embedded hardware, featuring a frontend interface with vanilla JavaScript, Google Blockly, and WebSocket integration for real-time device communication, and a Node.js backend for compiling and uploading block-based code.
• Created Coco AI Model Training Platform, a web application for training YOLO-based object detection models, complete with a frontend for dataset annotation, training visualization, and model export; implemented a RESTful Python backend (Flask) to orchestrate AWS EC2 jobs, automate model conversion for edge computing devices, and manage concurrent requests via Nginx reverse proxy.
• Engineered CocoPi Developer Tool, a React (with Redux) web tool for debugging embedded Linux boards via WebUSB, incorporating custom components for terminal access, file browsing, and comprehensive device control.
• Established and maintained Git-driven developer workflows, including a code deployment dashboard, changelog automation, and a real-time service status page, supporting robust CI/CD-style staging and production deployments.
• Managed, root-caused, and effectively resolved complex system-level issues, including inconsistent behavior on outdated PC setups and web service instability under load; led on-site debugging and deployed production-ready fixes.

Recognized with an Outstanding Short Paper Award for research on an agent-aided collaborative learning environment.

Co-authored a publication on a remote collaboration system, detailing an effective visual echo cancellation algorithm.