Turning every pitch into performance intelligence.
Founding Minds developed BullpenAI, a real-time baseball analytics system combining high-speed cameras, computer vision, and motion analysis to measure pitch trajectory, location, and pitching mechanics. It gives coaches and players objective information they can use during training and track across sessions.
The situation
Coaches evaluate pitches through observation and experience: watching, remembering, and writing notes afterward. Without a consistent digital record, it is difficult to replay a pitch or compare performance over time.
A coach might notice an inconsistent release, a change in delivery, or variation in ball location. Judging whether that pattern is improving across hundreds of pitches takes reliable measurement.
Instrumented analysis systems exist at the top of professional baseball. For college programs, academies, and amateur clubs, consistent pitch-level analysis can be harder to access and maintain.
What made it hard
Capturing fast motion with enough visual detail.
A pitch runs from wind-up to release in a fraction of a second, and the arm at release is where motion blur degrades an ordinary camera’s frame. The system uses a purpose-built, high-speed camera setup. The work was extracting measurements from fast-moving objects, not only recording video.
Tracking the ball and the pitcher.
Trajectory analysis and biomechanical analysis are separate computer vision problems with separate failure modes. The system tracks the ball’s movement while analyzing the pitcher’s motion.
Producing outputs coaches can trust.
The system calls strike or ball and reports which of nine zones the ball passed through. A coach can check these outputs against what they saw, so they needed to be dependable.
Making advanced analysis practical.
The goal was to make useful pitching analytics available beyond elite professional training environments. That placed constraints on camera hardware, computation, application design, and operational simplicity.
What was built
A high-speed camera rig captures the pitcher from wind-up through release, and the system analyzes each pitch automatically, in three parts.
Pitch capture and tracking
High-speed cameras, a deep learning model that detects the moment of the pitch, analysis of the ball’s flight to estimate trajectory, and a radar gun for velocity.
Pitch and movement analytics
Strike-or-ball classification, nine-zone pitch location, and biomechanical analysis that tracks body movement and surfaces inconsistencies in mechanics.
Player performance platform
A web application for real-time use during a session. Every pitch is stored for replay and historical comparison, and a player profile builds from the accumulated record for coaches, players, and talent scouts.
The system maps each pitch to a nine-zone grid, making pitch location measurable and comparable across sessions.
What changed
From individual observations to a measurable pitching history
BullpenAI turns recorded pitches into structured performance information, allowing coaches and players to review pitch location, velocity, trajectory, and pitching mechanics during training and compare results over time.
The engineering decision that mattered
BullpenAI required more than applying computer vision to recorded video. The team combined high-speed capture, ball tracking, biomechanical analysis, and a usable application to turn rapid physical events into repeatable measurements that coaches and players could interpret.