I’m building an AI-powered workflow to quantify how professional footballers perform across three pillars—physical attributes (speed, stamina), technical skills (dribbling, passing) and tactical awareness (positioning, decision-making). The idea is simple: feed the system raw match footage or event/tracking i have the source ready for system i need some one integrate all my HTML in one web site that do the following data and receive objective, match-by-match numbers I can compare over an entire season. Support Arabic and English The web work as platform for succor match analyses and suggest advice to cotch based on live streaming analyses and video upload paid service Second is admin ability to upload torments with expected results for teams based on AI analyses for marketing purpose generate list of teams based on total skills speed tallness factors. Third Players performance analyst from video tapes and live streaming report paid service Live chat support instant auto translation Arabic to English vise versa Distance cutch subscriptions page with comments on the live games and advices by cutch paid service Payments gateway I have many HTML pages used by admin for analyses User will not see the Ai functionalities only get results The core tasks involve: • Sourcing or accepting my existing Opta-style event logs and 1080p video files • Designing or fine-tuning a model (Python, TensorFlow/PyTorch, OpenCV, or a comparable stack) that automatically detects players, ball events and positional context frame-by-frame • Converting detections into metrics such as sprint count, top speed, distance covered, dribble success rate, progressive passes, heat-maps and expected threat added from positioning choices • Packaging the outputs in a clean CSV/JSON plus a lightweight dashboard (Streamlit or similar) that lets me filter by player, match period and metric Acceptance is straightforward: for a provided 90-minute test match, the dashboard must populate with all six metric categories and the raw CSV values must line up (±5 %) with manually sampled ground truth I’ll share. No fancy presentation required—just robust, well-commented code and clear setup instructions so I can reproduce results on my own GPU server.