ClearSite | Devpost

Architecture (open in new tab to see full)
Landing page
Logo

Project Inspiration

Artificial Intelligence has done great things for humanity. Hackathons like GenAI Genesis 2026 are a celebration of that—gatherings of builders and innovators across the world to create meaningful solutions to global issues.

Facilitating the massive userbases of flagship models like Google Gemini are data centers. In fact, global investment in data centers nearly doubled in just two years, hitting an astounding $500 billion in 2024 alone according to the International Energy Agency (IEA).

However, we ask that you take a step back to consider the following: Is this sustainable?

In 2024, data centers consumed roughly 415 terawatt-hours (TWh) of electricity; about 1.5% of total global demand. The IEA projects this will more than double to 945 TWh by 2030. Furthermore, a single large data center can consume up to 5 million gallons of water every single day. That is equivalent to the daily water usage of a town of 50,000 people. In the U.S., nearly half of the data center capacity is concentrated in just five regional clusters, causing local utility bills to soar and forcing the construction of new gas power plants just to handle the load.

Thus, a lot of people aren't very happy about AI. According to early 2026 research by Data Center Watch, local grassroots opposition delayed or canceled an astonishing $98 billion worth of data center projects between March and June 2025 alone.

As developers, we need to remember the average citizen is not excited about OpenClaw or Claude Code. They just want cheap energy and peaceful neighborhood. But data center proposals are filled with technical jargon, making it even harder for municipals and industry to reach a sustainable development plan together to address these concerns.

Therefore, we built ClearSite —Translating complex data center impacts into transparent, sustainable community choices.

Product Summary

ClearSite is an automated, AI-driven assessment platform designed to help municipalities, urban planners, and local citizens easily evaluate the environmental and socio-economic impacts of proposed data centers. By translating dense, highly technical proposals into clear, actionable insights, ClearSite empowers communities to make informed, sustainable development decisions.

How It Works & The User Experience

The user experience is designed to be frictionless and transparent:

Upload & Extract: A city planner or community representative uploads a raw data center proposal (e.g., a PDF) to the ClearSite web interface.
Automated Analysis: Behind the scenes, our system parses the document, extracting crucial metrics like IT load, Power Usage Effectiveness (PUE), and facility type.
Interactive Dashboard: The user is presented with a comprehensive dashboard featuring interactive maps (via Leaflet) and clear visual scorecards. The proposal is graded across multiple dimensions: Grid Strain, Water Usage, Carbon Footprint, and Community Fit.
Council Memo Generation: Finally, the user can instantly generate a jargon-free, policy-aligned "Council Memo," ready to be presented at town hall meetings or used in municipal planning discussions.

Addressing the Problem

Data center proposals are typically bogged down in technical jargon, making it incredibly difficult for local governments to accurately gauge how a facility will impact their local power grid, water supply, and community. ClearSite bridges this gap. It demystifies the technical parameters and provides objective, data-backed impact scores, helping to prevent resource crises and ease grassroots opposition by fostering transparent, sustainable planning.

Innovation & AI Utilization

To truly embody the spirit of sustainability, we built ClearSite to run 100% locally on a single CPU, entirely avoiding the carbon footprint of cloud-based AI inference APIs.

Extreme Efficiency w/ BitNet: We utilize 1.58-bit quantized LLMs running via Microsoft's BitNet framework directly on CPU. This handles the complex generative tasks—reading raw PDFs, structuring data, and writing memos—using a fraction of the energy required by traditional models.
Agentic Orchestration & Verification w/ Railtracks: To ensure zero hallucinations in policy or math, our LLM operates within the Railtracks framework. A dedicated "Verifier Agent" strictly audits the generated memos against deterministic mathematical calculations and policy databases, triggering automatic repairs if any fabricated numbers or clauses are detected.
Predictive Machine Learning:
- A CatBoost "Site Fit" model assesses geographic suitability by analyzing population density, water stress, and local carbon intensity (trained using a custom synthetic data generation pipeline).
- An XGBoost "Grid Strain" model predicts the exact probability that the new facility will push local power grids (like IESO in Ontario) past critical safety thresholds, factoring in historical demand, seasonality, and time of day.

By combining hyper-efficient local AI with rigorous environmental data pipelines, ClearSite proves that powerful, impactful generative AI solutions can be built responsibly, sustainably, and transparently.

Technology Stack

Languages: Typescript (frontend) + Python (backend)
Frameworks and Libraries:
- Railtracks: Agentic Orchestration & Verification
- BitNet: LLM Inference framework
- XGBoost + CatBoost: Local ML training
- Leaflet + OpenStreetMap: Map displays
- Next.js: Frontend
- FastAPI: Backend
Data Pipelines:
- Automated Ingestion: Scripts dynamically pull and compile public environmental and demographic datasets, including IESO (Ontario) hourly demand, StatCan census profiles, municipal water usage, and AAFC drought monitors into a local SQLite database.
- Proposal Normalization: Unstructured data center proposals uploaded as PDFs are parsed locally. We use agentic orchestration to extract raw text and normalize it into structured JSON parameters (e.g., IT load, PUE, facility type), gracefully falling back to deterministic regex parsing if needed.