Hands-on AI fluency
Building real systems with modern AI tools, end to end
A recruiting intelligence system Built solo in two weeks
For two weeks every spring, thousands of college basketball players enter the NCAA transfer portal to find new schools. It is a high-stakes and chaotic period. For Carnegie Mellon, it is the most consequential stretch of the year.
I built a system that identifies CMU's top recruiting targets, flags when they signal intent to transfer, and enables coaches to act in minutes instead of hours or days.
Created by a strategy & partnerships executive using modern AI tools.
What it does
→
→
→
Identify
AI evaluates 2,700+ players for academic and on-court fit
Monitor
Intelligently tracks player information across a fragmented D3 data landscape
Notify
SMS sent to the coaching staff within minutes of a qualifying player entering the portal
Inform
AI-generated player background dossiers, ready in ~5 minutes
What it produced
Top 10 Signing
Division 3 commitment
Top-10 ranked big man committed to the program
3 hrs. to 15 mins
Qualifying player awareness
Provides a first-mover advantage
2 hrs. to 5 mins
Player background brief generation
Frees the staff to act, not research
<$100
Total all-in cost
No vendors, no agency, no team
2 weeks
Idea to production
From first concept to live system
16
Python modules
Data collection, detection, alerting, and research
What this actually demonstrates
Basketball is the example. The skills are what I bring to the work.
Strategic analysis
Turning a vague situation into a sharp diagnosis
End-to-end ownership
Concept to live system in two weeks. Solo. Under $100
Rapid domain ramp-up
Operating in new domains. Same caliber of work.
01 · The Opportunity
why transfers, why now
My assignment was simple: help CMU get better, faster.
Carnegie Mellon men's basketball brought me on as a strategic consultant during a roster rebuild under a new coaching staff. As I analyzed the data, the same answer kept appearing: transfers.
The pattern was hard to ignore. Teams that used transfers were significantly more likely to win, but very few D3 programs pursued them.
The data showed that the transfer portal was one of D3's largest market inefficiencies. For CMU, it was the fastest way to accelerate the rebuild.
02 · The Challenge
why D3 adoption lagged
D3 Transfer Adoption Gap
My analysis revealed a stark contrast:
58%
6%
D1 roster building had moved to the portal. Selective D3 had barely adopted it.
D3 Constraints
D3 teams hadn't adopted the portal due to operational barriers
Limited staff + budget
Minimal time, budget, or personnel to dedicate
Player volume
Thousands of candidates to sort through
Academic filtering
Few players meet admissions standards
Fragmented information
Key signals live across disconnected sources
Missing infrastructure
D3 has no paid analytics, scouting, or recruiting platforms
The opportunity existed. The tools didn't.
So I built them for Carnegie Mellon.
03 · The Pipeline
four phases · end to end
Phase 01
Identify
Turn thousands of portal entrants into a qualified watchlist.
- Classifier model scores 1,300+ candidates by fit with CMU's returning roster
- Uses AI to group players by role, matching them to CMU's needs
- Filters for CMU academic fit before any outreach
Machine Learning Analysis
Perplexity
Google Colab
Phase 02
Monitor
Stay current on the full screened player universe across a fragmented D3 information landscape.
- Scans structured feeds, player socials, and open-web coverage
- No single source tracks D3 activity, so the system canvases all of them
- Surfaces portal entries, intent signals, and news coverage across the player set
Python
Scheduled jobs
Google Sheets
Phase 03
Alert
Push to the staff the moment a qualifier enters.
- SMS within minutes of portal entry
- No dashboard refresh, no morning catch-up
SMS
Email gateway
Phase 04
Intelligence
One click on a Priority List player fires a full background brief, generated on demand.
- o3-deep-research pulls biography, family, academic history, interests — anything publicly available about the player
- Brief ready in ~5 minutes — prep that used to take an afternoon
o3-deep-research
04 · Data Ingestion
sources · methods · use cases
05 · Beyond the Website
a content system sits alongside the tracker
Content System
Fully AI-generated recruit video
Every element — footage, music, narration — is AI-generated. Customizing a version for a specific recruit takes ~30 minutes; equivalent output from a professional crew would cost tens of thousands. Completed with no video production experience.
Why it matters
Detection + decision + outreach, complete
The tracker tells you who and when. The AI brief tells you whether to pursue. The content system gives you something specific to say when you reach out — a personalized video that shows the recruit you've done your homework.
06 · Built With
the toolkit behind the system
Stack ·
Claude (Cowork)
v0
ChatGPT
Perplexity
o3-deep-research
Twitter API
Google Colab
Google Sheets
Replit
07 · About
who built this
Bio
Scott Johnson most recently spent six years at Verizon in corporate strategy, serving as the strategic partnerships lead for the company's relationships with Disney and IBM. Before Verizon, he worked as an emerging technologies researcher at Harvard Business School.
Portal Monitoring
Active
Portal Entries
742 today
Priority List Adds
2 today
AI Player Briefs
2 generated
Coach Alerts
Sent 8:14 AM
| Player | School | Pos | Class | Player Archetype | Archive |
|---|
| Player | School | Pos | Class | Reason Filtered |
|---|