~/About~/Systems~/Foundry~/Blueprint~/Journal~/Projects
Book a Call

NBA Scenario Engine

The complete system record. Choose the proof lens that fits your role.

View on GitHub

Choose The Proof Lens

One system record, three reader paths.

Feature-frozen3-act recordGitHub available
Business proofFounder / operator

Turning Injury News Into Betting Edges Before the Lines Move

Star player ruled out on Twitter, sportsbook lines adjust within 10 minutes. This pipeline finds the teammate edges inside that window in under 3 minutes.

System FlowData ModelArchitecture LayersDecision Log
Open business proof6 min read
Architecture proofCTO / architect

Architectural Brief: NBA Scenario Engine

NBA game logs only exist for games a player appeared in. A two-stage model splits DNP classification from minutes regression to escape the survivor bias.

System TopologyInfrastructure DecisionsConstraintsDecision Log
Open architecture proof5 min read
Technical proofSenior engineer

Why I Split Minutes Prediction Into Two Models Instead of One

NBA game logs hide every DNP, so the regressor never learns zero-minute players and predicts 11 minutes for anyone ruled out. A two-stage model fixes it.

Code SampleFailureConstraintSurprise
Open technical proof7 min read

Related Systems

More proof in the same problem space.

View systems map

Multi-Agent RAG Platform

Retrieval and model routing

The optional knowledge graph essay shows where model ambition met memory cost and retrieval value.

Contract Lifecycle Engine

AI extraction with confirmation

Strongest example of keeping AI-generated obligations in Pending until a human confirms them.

Centralized Property Intelligence Hub

Searchable operational knowledge layer

Shows how replicated property data became a queryable intelligence layer without calling four services per question.