AI developer needed: cricket technique analysis from phone videos

About the project We’re building an AI coach for young cricketers, covering batting and bowling. A player films a shot or a delivery on a phone; the app tracks their body movement, scores their technique against a coach-written rulebook, and gives simple fixes and drills. We’re launching with cricket academies in India, and then we look to scale. The product is being built from scratch, and the AI scoring is the core of this contract. Phase 1: batting demo (this contract), about 4 weeks Upload or record a 3–6 second side-on clip of one batting shot (front-foot drive and forward defence) Detect the batter and track body keypoints frame by frame (MediaPipe, MoveNet, RTMPose or similar), robust to outdoor nets and budget Android phones Find the key moments of the shot (stance, front foot landing, contact) and calculate measurable checks (e.g. head over front knee, front foot to the line of the ball, balance) Score the shot out of 100 using rules our coaches provide, with sub-scores, one or two plain-English fixes, and a “please film again” response when confidence is low Show results over the video (skeleton overlay, key frames) in a simple Flutter app or responsive web app A simple labelling tool, or exported data, so coaches can mark clips and we can measure how often the AI agrees with them Test on our real academy footage (supplied after contract) and report accuracy against the coaches’ marks Design the system so bowling can be added in Phase 2 without rebuilding Phase 2, if Phase 1 goes well A full Flutter app for iOS and Android, with accounts, parental consent, a coach dashboard, Hindi and English feedback, cloud hosting in India, more batting shots, and bowling analysis: action, run-up, front-foot landing and release, and later ball speed and line and length. What we need from you Proven computer vision work with human pose estimation, ideally in sport or fitness. Please include links or videos Flutter or React Native experience; experience with on-device models (TensorFlow Lite, Core ML or MediaPipe) is a strong plus Experience processing video in the cloud (Google Cloud, AWS or Azure) Ball tracking or object tracking experience is a plus Clear communication in English, weekly demos, and code in our GitHub from day one Terms Fixed price for Phase 1, paid against milestones All code, models, data and designs are owned by our UK company, with full IP assignment (worldwide and perpetual) and an NDA Footage of young players must be handled securely, stored only as agreed, and deleted at the end of the contract. No reuse of our footage, rules or models for anyone else To apply, please answer: Link to a pose-estimation or sports-analysis project you built, and your role in it Which pose model would you use for a side-on batting clip on a budget Android phone, and why? On-device or cloud processing for this demo? Briefly explain your choice Fixed price and timeline for Phase 1, and how you would split the milestones Who would do the work: you alone, or a team? Where are you based? Have you built ball tracking or bowling-action analysis? How would you measure ball speed from a single phone video?