Scout

College Sports Scouting for the year 2026.

Comment

Inspiration

Gunnar is a coach and former pro lacrosse player; Jake is an AI engineer passionate about how computer vision can empower athletes. Together, they set out to democratize college recruiting.

Currently, scouting relies heavily on expensive camps and curated highlight reels. Scout’s mission is to level the playing field, turning recruiting into a meritocracy where hard work and growth are as visible as height and weight. We help scouts find "diamonds in the rough" and give every athlete a fair shot at the collegiate level.

What it does

Scout researches athletes, grades position-specific combine footage, provides AI-driven coaching tips, and streamlines the discovery process for recruiters.

Unlike "Web2" recruiting sites that place the entire burden of presentation on the athlete, Scout actively helps athletes grow. We offer dynamic grading of drills that goes beyond traditional metrics. For scouts, the platform replaces complex filters and endless search pages with a natural language interface. You can talk to Scout like you would talk to a coach to find the exact player your program needs in half the time.

For scouts, this makes it simple to find athletes. You can talk to scout like you would talk to a coach, no more complex filtering or pages of results. Find the exact type of kid your program needs in half the time.

How we built it

Next.js: Powering the web app and API routes in a unified codebase.

Firebase/Firestore: Managing user profiles, events, reports, and metadata.

Google Cloud Storage: Hosting raw high-resolution video uploads.

Gemini AI: Analyzing drill videos, generating coaching/scout reports, and conducting background research.

Challenges we ran into

The biggest hurdle was differentiation: why would a scout or athlete leave an established platform? We identified three primary pain points to solve:

Profile Fatigue: Manual data entry is tedious. We built a Research Agent that automatically updates profiles with new event data, removing the administrative burden from the athlete.

The "Invisible" Athlete: Great players often lack professional highlight reels or access to top academies. We used Gemini’s multimodal capabilities to grade raw drill footage, extracting "overlooked" abilities that don't show up in a standard box score.

Search Friction: Scouts are busy. Instead of toggling endless filters, we implemented natural language search. Whether a coach needs "a left tackle with long arms" or "a speedster for faceoffs," Scout finds them instantly.

Accomplishments that we're proud of

Multi-Agent Architecture: Deployed 5 specialized agents using a mix of Gemini 1.5 Flash and Pro models.

High-Accuracy Search: Our agent accurately locates athlete links with minimal hallucination.

Computer Vision Precision: Our video agent successfully automates grading for 20-yard dashes, shuttle times, and wall-ball counts.

Smart Querying: Developed a query-writing system specifically trained on our database schema.

What we learned

We discovered that the Gemini 3.0 Pro model is superior for long, complex movements (like tracking a lacrosse stick during wall-ball drills).

We also tackled the logic of AI data management: determining when it is relevant to regenerate reports and when to allow the AI to write to the database. We learned that rigorous testing is required to ensure AI acts as a supportive tool rather than a "black box" that overwrites trusted user data.

What's next for Scout

We launched with three drills for one sport, but the architecture is built to scale. We plan to:

Expand Drills & Sports: Implement sub-agents optimized for hundreds of different athletic movements.

API Integration: Connect to public athletic APIs to make Scout the "single source of truth" for multi-state recruiting, saving scouts from scouring dozens of disconnected databases.

Share this project:

Updates