Opening page…
Projects
School projects, work experience, product practice and personal builds.
From academic work to real-world practice, each project marks a different stage of learning.
36 projects found
Start with roles and evidence, then explore the full archive.
Jingbian Tutoring is a tutoring service platform that connects families of primary and secondary school students with qualified university student tutors. The project aims to reduce information asymmetry and help families discover suitable tutors more efficiently.
Education Technology · Tutoring Marketplace · University Student Tutors
A conversational AI Agent designed around Lalamove’s urban freight-ordering scenario. It understands natural-language or voice requests, extracts order parameters, clarifies missing information, recommends suitable vehicle types, and generates a structured order draft for user confirmation.
An approximately 80% one-shot confirmation rate; · Internal testing; methodology pending
Atoms, formerly MetaGPT X (MGX), is a multi-agent AI Coding platform developed by DeepWisdom. It transforms natural-language requirements into functional applications through the collaboration of AI product managers, engineers, architects, and data analysts. As a Full-Stack AI Product Manager Intern, I worked on Builder Runtime and AI Infrastructure capabilities—including Secrets, Auth, AI Capability, Cloud & AI Balance, Overview, Preview, Restore, and Publish—to simplify backend configuration, model integration, cloud-resource management, and application deployment for non-technical builders.
Multi-Agent System · AI Coding · AI Application Builder
A tractor-mounted embedded system that enables farmers to capture field observations through voice interaction, automatically associate recordings with GPS and task information, and generate structured MP3 and JSON records for later analysis.
Improved overall system reliability by 40%
Exploring technology through coursework and research.
A tractor-mounted embedded system that enables farmers to capture field observations through voice interaction, automatically associate recordings with GPS and task information, and generate structured MP3 and JSON records for later analysis.
Improved overall system reliability by 40%
An interactive digital hourglass using a 64×64 RGB LED matrix, Raspberry Pi, and LIS3DH accelerometer to simulate gravity-driven sand flow, countdown timers, and motion-responsive visual effects.
Developed a digital hourglass using a 64×64 RGB LED matrix
A provincial-level innovation project integrating artificial intelligence, IoT, edge computing, and cloud technologies to support real-time EV charging monitoring, predictive maintenance, fault detection, and dynamic load balancing.
Reduced system failure rates by 25%
A document image analysis system combining PaddleOCR, DBNet, Mask R-CNN, and LayoutParser to recognize text and reconstruct structural elements from distorted, low-resolution, and multi-layout scanned pages.
Achieved 96% OCR precision
An individual research project investigating machine learning-based driver fatigue detection through multimodal physiological and behavioral data, including facial expressions, eye movements, head motion, EEG signals, and driving behavior.
Achieved 99.5% processed-data integrity
A national research project integrating non-contact ECG acquisition, environmental sensing, multi-sensor data fusion, and real-time alerts to identify driver fatigue and hazardous in-vehicle conditions.
Maintained environmental parameters within 5% of optimal conditions
A Python-based real-time data collection system using Wi-Fi and TCP communication, integrating error correction, deep clustering, frequency-domain analysis, and optimized buffer management.
Achieved 99.5% data transmission reliability
An integrated robotics course project at Cornell ECE 5772, combining multi-sensor perception, probabilistic localization, occupancy-grid mapping, and autonomous path planning on a mobile robot platform.
SLAM · Mapping · Localization
A graduate-level embedded systems course at Cornell University focused on developing hardware–software integrated applications using Raspberry Pi 4B, Linux, Python, and C++.
Raspberry Pi application development
An embedded calculator developed with the TM4C123GH6PM microcontroller, supporting numerical input, arithmetic operations, result calculation, and real-time display through a matrix keypad and display module.
The final system implemented a complete calculation process from keypad input to result display. It supported numerical input, basic
Isotope is a short narrative set on a lunar helium-3 mining base, following worker Sam Bell as he discovers a clone of himself and confronts questions of identity, human rights, and corporate disregard for individual life.
A Unit 4 FPGA assessment that implements a stopwatch on the MAX10 architecture in Quartus, covering configurable logic, timing modules, and a written design report of the hardware implementation.
Through the process of designing, coding, and testing each of these modules, and integrating them into a complete system, I have deepened my understanding of digital system design with FPGAs. I have learned how to handle clock signals, deve…
An eight-week embedded mini-project that ports Space Invaders to the STM32 Nucleo-L476RG, using a joystick and LCD as the core interface and optional peripherals such as LEDs, switches, or an accelerometer.
A presentation on photovoltaic technology in Malawi, contrasting thin-film cells with conventional silicon panels and discussing cost, climate constraints, and how PV could complement unreliable hydropower in a country with limited electricity access.
A semiconductor-technology presentation covering visible-light Li-Fi communications with GaN LEDs, and terahertz research including quantum-cascade lasers, their advantages, limits, and potential applications.
A lab-pack app for engineering beginners, especially SWJTU-Leeds students. It includes a matrix solver for circuit analysis, a filter-component calculator, and Buck/Boost converter design tools to cut calculation time and errors in lab sessions.
A washing machine control panel developed with the STM32L476RG Nucleo-64 development board, integrating buttons, LEDs, sensors, a timer display, a thermistor, a buzzer, and a serial status interface.
STM32 microcontroller development
A remote-controlled buggy built around the M6513 motor, a Bluetooth module, and an H-bridge driver prototyped on a breadboard then soldered to a board. A metal tank turret adds 360-degree rotation through a secondary motor and gear train.
A laboratory exercise on DC motor principles: winding a coil, assembling rotor and stator, and building drive circuits with a transistor, Hall sensor or reed switch, magnets, and wire to practice circuit debug and measurement.
An academic presentation examining the importance of innovation in engineering and evaluating a discipline-specific innovation from sustainability and engineering ethics perspectives.
Engineering Innovation · Contribution to Sustainability · Ethical Consideration
An introductory engineering project involving the design, construction, and experimental testing of a balsa wood glider to improve its flight distance and stability.
The project provided an initial understanding of aerodynamics, weight distribution, and lightweight structural design. It also introduced the fundamental engineering process of designing, building, testing, and iterating a physical prototyp…
Projects built through contests and challenges.
P.A.T.H. is an integrated in-vehicle health management system that continuously monitors drivers’ physiological conditions and uses intelligent algorithms to assess long-term fatigue and psychological risks, helping transportation companies improve operational safety and employee health management.
The project developed an integrated, non-wearable in-vehicle health monitoring solution covering physiological data collection, environmental sensing, intelligent analysis, and enterprise health management. The prototype monitors ECG signal…
Ruiyuan is a real-time health monitoring and inspection system that uses vehicle operating data, big data analytics, and artificial intelligence algorithms to evaluate the battery, motor, and drivetrain systems of new energy vehicles.
Through multi-sensor data fusion and deep-learning-based analysis, the project improved abnormal-condition detection accuracy by approximately 27% and reduced false alarms by approximately 32%. The system also enabled potential battery safe…
HandsFree is an intelligent loading and unloading solution that integrates automated cargo-handling equipment, a control system, and an information management platform. Designed for unmanned delivery vehicles across community, warehouse-to-store, and factory logistics scenarios, it enables efficient, safe, and fully autonomous cargo handling, helping autonomous delivery vehicles achieve true end-to-end unmanned operation.
The project completed the initial development and preliminary debugging of the HandsFree intelligent loading and unloading device and its control system. The prototype demonstrated the technical feasibility of integrating an intelligent car…
Wireless Stream is a compact and efficient watt-level wireless power transfer system designed for tower-side sensing and monitoring equipment. By integrating multi-stage magnetic resonance coupling with high-voltage insulator structures, the project provides reliable and continuous power for cameras, wireless concentrators, sensors, and bird-deterrent devices deployed along transmission and distribution lines.
Energy · Hardware · Grid
Introduced sensors, automation and intelligent algorithms into pigeon breeding to improve loft management.
IoT · Automation · Agriculture
Internships and roles in real product and enterprise settings.
A record of my journey in the Alibaba Cloud Management Trainee Program, covering cloud computing foundations, business and customer perspectives, cross-functional collaboration, AI workflow experimentation, and my exploration of the Solution Architect career path.
Alibaba Cloud · Management Trainee · Cloud Computing
Atoms, formerly MetaGPT X (MGX), is a multi-agent AI Coding platform developed by DeepWisdom. It transforms natural-language requirements into functional applications through the collaboration of AI product managers, engineers, architects, and data analysts. As a Full-Stack AI Product Manager Intern, I worked on Builder Runtime and AI Infrastructure capabilities—including Secrets, Auth, AI Capability, Cloud & AI Balance, Overview, Preview, Restore, and Publish—to simplify backend configuration, model integration, cloud-resource management, and application deployment for non-technical builders.
Multi-Agent System · AI Coding · AI Application Builder
During an internship with the AI computing product line of a leading ICT technology company, I participated in solution exercises covering an enterprise Agent platform for telecom marketing, AI computing resource planning, healthcare AI solution presentation, securities-industry bidding simulation, and AI-powered vocational education research.
Product Marketing · AI Infrastructure · Industry Solutions
A conversational AI Agent designed around Lalamove’s urban freight-ordering scenario. It understands natural-language or voice requests, extracts order parameters, clarifies missing information, recommends suitable vehicle types, and generates a structured order draft for user confirmation.
An approximately 80% one-shot confirmation rate; · Internal testing; methodology pending
A configurable internal platform connecting registration, team formation, topic selection, submissions and judging.
The report records a 20-day development period, August 3–22, 2026.
A shared live state for guest profiles, schedules, vehicles and reception coordination.
The report records concentrated development from September 11–21, followed by October refinements. Concurrent overwrites, stale clients clearing fields and mock startup flicker exposed weaknesses in the early write model. Incremental update…
Internal cross-cultural sales practice through AI personas, voice conversations and feedback.
The report records 14 concentrated commits, 39 base personas, 12 quests and eight TTS voices. Speaking is the implemented 50-point scoring component; the Thinking judge-crm endpoint is unimplemented, so the nominal 100-point framework is no…
An observable, retryable 16-stage workflow transforming stories, characters and photo inputs into video.
The report demonstrates a 25.2-second, 1080P, 30FPS output with 16 visible stages. Script, continuity and final-film checks are separate; a single sample score is not a general success rate. Unavailable music services are skipped, and const…
Trying to turn ideas into something people can use.
Jingbian Tutoring is a tutoring service platform that connects families of primary and secondary school students with qualified university student tutors. The project aims to reduce information asymmetry and help families discover suitable tutors more efficiently.
Education Technology · Tutoring Marketplace · University Student Tutors
Paisheng Competitions was a university student competition training platform I founded during my undergraduate studies. Covering more than 50 recognized competition categories, the platform supported students in innovation and entrepreneurship, art and design, computer science, electronic design, and other discipline-specific competitions.
Competition outcomes are influenced by project foundations, team commitment, competition intensity, and judging standards. The platform needed to support strong results while preserving genuine student development.