1. Developed the backend of a low-latency, plugin-free surveillance web viewer by requesting Real Time Streaming Protocol (RTSP) streams from a Network Video Recorder (NVR) and transferring H.264 NAL units to the frontend via WebRTC. 2. Created a surveillance video playback service backend using Hikvision NVR’s C API to query date information and request Transport Stream (TS) packets; integrated C code into Golang with cgo and provided streams using HTTP Live Streaming (HLS). 3. Implemented communication protocols (Modbus, YDN23) in Golang to fetch and write data on industrial electronic devices. 4. Built simulated devices for testing and demonstrating our IoT web system. 5. Implemented a Docker web UI that could be easily used by our clients by using Golang Docker client API for the backend. 6. Created a simple CI/CD pipeline using GitLab CI/CD and kaniko, including Golang code testing, Docker image building and pushing.

1. Built a Machine Learning model to classify images of semi-finished products as either OK or NG by training a convolutional neural network with data augmentation on thousands of images. 2. Developed a simple but effective segmentation process to segment images of semi-finished products into different regions by using Template Matching. 3. Integrated YOLO’s pre-trained person detection model into a Person Re-identification system to build an intelligent surveillance system. 4. Built a people counter using Kinect, NVIDIA Jetson TX1 and the open-source project OpenPTrack for pedestrian tracking. 5. Wrote and maintained a Dockerfile for Machine Learning model development and deployment. 6. Utilized gRPC to communicate between Python and C# programs.