Independent prototype
AdTech Decision Intelligence prototype
An independent project exploring how AI can interpret public data while keeping calculations, evidence and uncertainty visible.
The product question
How can AI help a reader interpret multiple advertising and attention signals without losing sight of the evidence behind a conclusion? I use public data around the 2025 New York City mayoral election as a demonstration setting.
The approach
The prototype separates deterministic Python calculations from local AI interpretation. Its design uses a structured evidence bundle, checks on generated conclusions and a Next.js presentation layer. The project documentation names NYC Campaign Finance Board, NYC Board of Elections and Wikimedia pageviews as sources.
The product judgment
The signals measure different things. Wikipedia pageviews reflect page attention, not voter intent; advertising spending does not establish advertising effectiveness. The design aims to make those distinctions understandable and keep the underlying evidence accessible alongside the generated interpretation.
Project status
This is an independent prototype exploring analytical processing, local AI models and product presentation. It is separate from my professional delivery work at SMI. The emphasis is on making an interpretation traceable to evidence and communicating uncertainty clearly.