Project 02 · Football Intelligence
MatchMind AI
Turns one uploaded match video into a complete tactical report: formations, pressing, player performances and coaching recommendations.
Scene 1 / 6
One match video in.
Upload a single full-match video. No manual tagging, no spreadsheets.
Scene 2 / 6
The AI watches the whole match.
A multimodal model watches the whole match as video, start to finish.
Scene 3 / 6
It reads the shape.
Formations: how each team lines up and how that shape shifts.
Scene 4 / 6
It sees the press.
Pressing: who closes down the ball, how high, and how fast.
Scene 5 / 6
Every player, assessed.
Individual performances for the players on the pitch.
Scene 6 / 6
Coaches get the report.
A full tactical report with coaching recommendations, in about 2.5 minutes on average.
MatchMind · tactical report
Formations
Pressing
Player performances
Coaching recommendations
~2.5 min average analysis
The details
Under the hood.
MatchMind AI is a full-stack web app that turns a single uploaded match video into a complete tactical report (formations, pressing, individual player performances and coaching recommendations), powered by a multimodal AI model that watches the whole match as video. It is deployed and live. Separately prototyped an offline detection-and-tracking pipeline (RF-DETR fine-tuned on SoccerNet-Tracking, with a custom tracking layer) as research into running without external AI APIs.
- ~2.5 min average analysis time
- Automatic 3-key failover
- Streaming uploads, no size crash
Free hosting: the demo can take about 40 seconds to wake up.
Built with
Next project
HelixForge
Computational Biology