Projects

SmartAC

Constructed an IoT system in Python to dynamically adjust multiple fan speeds based on detected human presence and respective temperatures.
Integrated MLX90640 thermal camera with Raspberry Pi for thermal data acquisition by leveraging I2C communication.
Modified YOLOv5 object detection model to detect humans and display temperatures in video streams.
Combined thermal data with detection results to dynamically adjust fan speeds via GPIO-controlled PWM signals.
Designed a multi-threaded TCP server to receive and process fan speed data in real-time.
Accomplished real-time processing with less than 100ms latency, improving cooling efficiency and user comfort

Enhanced low-resolution images through a three-stage process: image pre-processing with singular value decomposition (SVD), resolution enhancement using Transformer modules, and final image fusion
Demonstrated a 0.02 PSNR improvement over leading models with only $1500 funding

Predicted user ratings on movies by utilizing the ensemble method, including user-based collaborative filtering, item-based collaborative filtering, and neural networks
Reduced MAE to 0.744, ranked 1st in all classes this quarter

Developed a fault-tolerant, distributed queue inspired by Kafka’s architecture, implementing leader election, replication, and partitioned topics for horizontal scalability
Achieved stable message throughput of approximately 1,500 items/second in a test environment with three brokers and three partitions, maintaining end-to-end latencies under 100 ms for over 90% of messages
Demonstrated system reliability through graceful handling of broker failures and automatic leader re-elections

Wrote 3000+ lines of C++ code to design a 64-bit compiler for Simple C
Implemented key stages of the compiler, including lexical analysis, syntax analysis, semantic analysis, storage allocation, and x86 code generation

Developed a full stack website allowing users to upload, search, and delete cosplayers’ photos
Implemented a various amount of services to provide users CRUD functions

Developed an anonymous platform within one day for users to login and publish anonymous posts