TIDAL WAVE

• 

  Logo

Team Members: David Wan, Shreyaan Chhbara, Aviraj Khanuja
Team Name: TIDAL WAVE
GitHub Repository Link: https://github.com/davdwan21/TIDAL-WAVE
Selected Tracks: Data Analytics, Cloud
Selected Challenges (Optional): Best Use of AWS Services, Best Use of NVIDIA Brev, Most Viral Idea, Most Innovative Idea, Best Use of Scripps Data

In ecological science, most systems assist data analysts by processing and visualizing data for retrospective analysis. However, this approach not only restricts the ability of non-experts to contribute to environmental conservation but also forces analysts to rely on educated guesses when predicting future outcomes. Our idea was simple: why limit everyone to the past? TIDAL WAVE is a data-driven simulation tool that allows users to act as policymakers by inputting environmental policies in natural language. The system translates these policies into global parameters and simulates how a coastal ecosystem responds over time.

TIDAL WAVE consists of several core components:
Translating natural language into concrete parameters
In order to turn the user's policy idea into environment parameters, we employ text-to-data generation with Groq's API. To ensure real-world accuracy, we developed a script that analyzed the oceanic data from CalCOFI, identifying 6 important parameters along with their medians and their average distributions. We then backtested the model outputs according to these distributions, ensuring that the simulation parameters returned by the model reflected real-world conditions.

Simulating an Ecosystem
The simulation itself operates on 6 global parameters and 7 different species. Each species is represented by an AI-agent which acts on behalf of the species. During each simulated year, agents decide on species-wide actions based on global parameters and environmental data. The ecological system then evolves according to these decisions, using established ecological population dynamics models. Equations and frameworks (such as Lotka-Volterra predator-prey dynamics and trophic coupling) calculate population interactions and adaptations, ensuring simulation accuracy and minimizing AI-generated hallucinations.

Visualizing the Results
We built a website using React, Vite, and Tailwind CSS to visualize simulation results. Users can input policy ideas through the interface, and TIDAL WAVE runs the simulation and displays the outcomes. Users can navigate through yearly data, observing changes in global parameters, environmental values, and population dynamics over time. By presenting predictions in an accessible format, users can make informed analyses of how their policy ideas affect the ecosystem.

Scaling our Project with AWS
We implemented a deployment system using AWS and Docker that runs simulations on EC2 and leverages ECR, ECS, and S3 for storage infrastructure. By scaling our storage and simulation capabilities with AWS, we can expand predictions by orders of magnitude—incorporating many more species, simulating larger geographical areas beyond the Southern California coast, and extending timeframes beyond five years. Additionally, EC2's parallel computing capabilities enable us to run multiple simulations simultaneously, facilitating direct policy comparisons. We've begun implementing this through AWS cloud deployment, though it remains in early stages due to hackathon timeframe and API credit constraints.

Built With

• ai-agents
• amazon-web-services
• brev
• groq
• python
• react
• tailwind-css
• vite