CoachAI
I coached 200+ athletes and spent years in biomechanics labs. Now I'm teaching AI everything I know. Pitching instruction, pitch design, and arsenal development -- built on Driveline methodology, not scraped from the internet.
Where it comes from
The coaching frameworks inside CoachAI aren't pulled from the internet. They're documented from real athlete work -- Rapsodo sessions, Trackman data, Blast Motion sensors, high-speed video, and biomechanics labs. I ran programming at 802 Baseball Lab and worked at Driveline Academy Flex. This is what I actually taught.
The Statcast integration connects that coaching knowledge to live MLB data. Benchmark pitch shapes against major league arsenals. Research tunneling strategies. Compare movement profiles across the full population of active pitchers. It's where domain expertise meets real data.
Capabilities (v1)
Pitching Instruction
Mechanics analysis, arm path, hip-to-shoulder separation, extension. Driveline-informed cues and drill programming based on the athlete's movement patterns.
Pitch Design
Arsenal building from first principles. Pitch tunneling analysis, grip adjustments, movement profile targets. Backed by Statcast shapes and expected outcomes.
Biomechanics Analysis
Kinematic sequencing, force application, injury risk assessment. Frameworks built from professional coaching experience with Rapsodo and Trackman.
Statcast Integration
Live MLB Statcast data via 24-tool MCP integration. Spin rate benchmarks, movement profiles, pitch usage trends across the league.
Arsenal Documentation
Per-pitch documentation: 4-seam fastball, kick-change, curveball, slider. Velocity targets, shape targets, usage strategy, and in-game attack patterns.
Status
v1 is an internal research tool. A structured knowledge base for pitching coaching and arsenal design that I actually use for real decisions. It works. It's not pretty yet.
v2 is in planning. The goal is a public coaching interface where any pitcher can get Driveline-caliber feedback from a system trained on years of real athlete data. That's the direction. Just rip it -- ship the v1, iterate toward the product.