Mediapipe Post

We're building a web-based system that analyzes human pose landmark data (MediaPipe-style 2D/3D keypoints) from video and detects structured movement events from noisy time-series signals. We need an experienced developer with strong applied computer vision and signal processing skills to improve event detection accuracy, eliminate duplicate detections caused by multi-peak motion, and design robust, adaptive logic for identifying distinct movement phases from normalized X/Y/Z coordinates.
Ideal candidates have hands-on experience with pose pipelines (MediaPipe, OpenPose, similar), temporal smoothing, velocity/acceleration modeling, peak detection, hysteresis, and event segmentation. This is not a frontend task — it requires someone comfortable working with real-world noisy motion data and building production-ready detection logic in TypeScript or Python. Please include relevant motion/pose-based work and briefly describe how you would detect a jump-like movement from ankle Y-coordin...