GridMark

Inspiration

Infrastructure projects like electrical grid expansions often face public resistance—not due to their utility, but because stakeholders feel their specific concerns aren’t addressed. We were inspired by political campaign tactics and modern A/B testing methodologies used in advertising and applied them to a neglected domain: localized infrastructure communications.

What it does

GridMark is a marketing simulation and testing platform for grid infrastructure projects that:

Accepts location-based inputs and project details
Aggregates GIS, census, and psychographic data
Simulates A/B tested campaigns (e.g. speeches, tweets, PDFs) across different demographic personas
Predicts reception and success scores by region and group
Displays results on an interactive GIS map with zoom-based granularity
Recommends how to improve targeting (tone, format, speaker identity, etc.)

How we built it

We built a prototype with LLMs in React with mock data. A real solution would use:

GIS & Environmental Data: Hard-coded datasets (e.g., topography, proximity to wildlife/farmlands) Demographic Analysis Engine: Uses U.S. Census data and synthesized psychographics to create local stakeholder profiles LLM-Driven Simulation Engine: Uses fine-tuned language models to simulate how different personas interpret and react to various media types A/B Testing Framework: Simulates how variations (e.g., post wording, speaker identity) perform across demographics

Challenges we ran into

None, with mock data.

Real challenges would mostly be accuracy in terms of synthesizing mock demographic personalities, and the actual simulated testing. Aggregating real data may be challenging but is definitely doable.