Computer Vision Engineer Needed for Real-Time Video Tracking & Counting System

I am looking for a Computer Vision engineer to help us build a small but reliable video analytics system for a fixed camera setup. The main goal is straightforward: detect and track objects moving through a defined area and produce reliable counts and timestamps. This will initially run on recorded video, but the implementation should be structured so we can move it to an RTSP camera stream later. What I need: * Object detection on video frames * Multi-object tracking across frames * Line/zone-based counting * Entry/exit direction detection * Timestamped detection records * Confidence scores * Basic handling of missed detections and duplicate counts * CSV/JSON output for downstream integration * Simple configuration for camera/source, detection classes, confidence threshold, and counting zones I’m not looking for a research project or a complicated dashboard at this stage. The priority is getting a clean, maintainable CV pipeline that performs consistently on real-world footage. Preferred approach: You can use YOLO or another suitable detector, together with ByteTrack, BoT-SORT, DeepSORT, or a better approach if you have a reason for it. OpenCV/Python is fine. The system should be designed with inference performance in mind, since the next phase will involve processing live camera streams. Deliverables: 1. Working Python CV pipeline 2. Detection + tracking + counting 3. Configurable zones/lines 4. Sample output data 5. Short setup/run documentation 6. Explanation of the main design decisions Please include **one example of a similar tracking/counting system you have actually built**, especially if it involved fixed cameras, RTSP streams, production video, or counting objects over time. I’m more interested in something practical and reliable than an impressive-looking demo.