Primary Explainer
Scroll down halfway if you're proudly kind of a tech geek.
Imagine a hospital emergency room where 50 people walk in at once. The doctors don't just help the person who screams the loudest; they help the person who is in the most danger first. SoilTok is an ER for the Earth.
Inspiration
Imagine the Earth is hurtling toward a massive "Game Over" screen because the soil is dying. Right now, when people try to fix the environment, they often guess where to help. They might pick a spot because it’s famous or because someone on TikTok complained about it. We wanted to build a "Smart Map" that stops the guessing and tells us exactly where the Earth is "bleeding" the most so we can save it before it's too late.
What it does
SoilTok is like a "Health Monitor" for the ground. It looks at different parts of the woods or farms and gives them a grade.
- Red means "Emergency: Fix this today or it’s gone forever."
- Yellow means "Keep an eye on it."
- Green means "Doing great." It helps leaders spend their money and time on the places that actually need help, instead of wasting it on places that are already doing fine.
How we built it
We combined three "Superpowers" to get the full story:
- Eyes in the Sky: We used Satellites to see the big picture (like how brown or green the plants are).
- Eyes on the Ground: We made a way for regular people to use their phones to report what they see right in front of them.
- Old-School Wisdom: We listened to people who have lived on the land for hundreds of years (Indigenous groups). They know things that a computer sensor might miss!
Technical Stack:
- Frontend: React 19, Vite, Tailwind CSS 4.
- Mapping: Leaflet.js with H3-js for spatial indexing.
- Icons & Animations: Lucide-react and Motion.
Challenges we ran into
The hardest part was getting everyone to agree. Sometimes the Satellite says "The ground looks dry!" but a person standing there says "Actually, I just saw a new sprout!" We had to write a special code (an algorithm) that acts like a judge to decide which information is the most trustworthy so the map stays accurate.
Accomplishments that we're proud of
We are really proud that our app actually listens to people, not just machines. Most science projects only care about high-tech sensors. We proved that if you combine high-tech satellites with the smarts of local people, you get a much better "Safety Grade" for the planet.
What we learned
We learned that "more data" isn't always better—"smarter data" is. We also learned that the people who live on the land usually know it better than any robot. We realized that if we want to save the world, we have to listen to the people who actually live in the areas we’re trying to fix.
What's next for SoilTok
We want to give SoilTok a "Crystal Ball." We want to use AI to predict the future. Instead of just seeing where the soil is dying now, we want the app to tell us: "Hey, if you don't help this forest in the next three months, it’s going to turn into a desert." We want to stop the problem before it even starts!
Inspiration
Current conservation efforts suffer from a "data-rich, insight-poor" paradox. We noticed that while satellite imagery and soil sensors are more accessible than ever, the actual allocation of resources—funding, labor, and equipment—is still largely driven by whoever has the loudest voice or the most recent (but often incomplete) report. We were inspired to build a system that acts as a triage engine, turning fragmented environmental signals into a clear, prioritized work order for the people on the ground.
What it does
SoilTok is a decision-support tool that ranks ecosystem restoration urgency. It moves beyond static heatmaps to provide a dynamic "triage score" for specific land parcels. By analyzing cross-sections of data, it identifies "Breaking Points"—areas where a small intervention today prevents a total ecological write-off tomorrow. It doesn’t just show you where the land is degraded; it tells you where your next dollar of investment will have the highest ROI for the planet.
How we built it
We built a weighted intelligence layer that synthesizes three distinct data streams:
- Macro-Scale: Integrated satellite APIs to track vegetation indices and moisture levels.
- Ground-Truth: Developed a pipeline for "Citizen Science" inputs, allowing field agents to upload localized observations.
- Deep Intelligence: Created a framework to treat Indigenous and local knowledge as a first-class data signal, providing the historical context that sensors often miss. The frontend was crafted as a high-performance HTML/CSS walkthrough to demonstrate the "scroll-to-reveal" logic of the triage process, emphasizing transparency in how scores are calculated.
Challenges we ran into
The primary hurdle was Data Normalization. Satellite data, text-based field reports, and qualitative historical anecdotes don't naturally speak the same language. We had to develop a logic gate that could weigh these inputs based on reliability—ensuring that a single outlier (like a sensor error) wouldn't trigger a false "Critical" alert, while still allowing urgent human observations to escalate the priority.
Accomplishments that we're proud of
We are particularly proud of our Signal Fusion approach. Most platforms choose between being "high-tech" (purely satellite) or "community-led." We successfully built an architecture that does both. Seeing the system correctly identify a high-risk hex-cell by validating a "Moderate" satellite signal against a "High-Urgency" local report was a major "aha" moment for the team.
What we learned
Building SoilTok taught us that restoration is as much about people as it is about soil. We learned that data volume is a vanity metric; what actually matters to decision-makers is Decision Quality. We also gained deep respect for the precision of Indigenous knowledge, which often predicted soil shifts days before the satellite data showed a measurable change.
What's next for SoilTok
The next phase is moving from a prototype to a live Predictive Triage model. We plan to:
- Integrate Anthropic's newest model of Claude AI: To move from reactive triage to predictive forecasting—identifying areas likely to reach a breaking point 3-6 months in advance.
- API Expansion: Creating a "Restoration API" so existing NGOs and government agencies can plug their own data into our weighting engine.
- Closing the Loop: Developing a "Post-Intervention" tracking layer to prove the impact of the decisions made through the platform.
Built With
- leaflet.js
- lucide
- motion
- python
- react
- tailwind
Log in or sign up for Devpost to join the conversation.