Project Eden

Inspiration

After researching more details about Biosphere 2, we found that there are useful data and interesting findings to be interpreted using the LLM's power from this challenge. We wanted our project to reflect the level of research we did, and decided to focus on the Desert Biome and the idea of monitoring competing species in a controlled environment.

What it does

Our LLM model uses data input given to us by the sponsors and interprets that data to the best of its abilities given a prompt. It is specialized to the Desert Biome right now.

How we built it

We build this using Ollama model qwen2.5 as our baseline and created a vectorspace using the data provided to produce a AI chatbot that answers questions about the desert biome.

Challenges we ran into

We had trouble training our qwen2.5 model because we didn't have enough data to do rigorous training. Our synthetic data had to be renewed multiple times to fit our constraints.

Accomplishments that we're proud of

We're happy with the quality of the synthetic data we created, the level of accuracy of the trained model and the ability of our team to work efficiently throughout the process.

What we learned

We learned a lot about setting up the environments for python and ollama due to many many mishaps throughout the training and setup. We also learned that with more specific prompts and longer conversations, generative AI will give responses of higher quality.

What's next for Project Eden

By feeding the model with more data and increasing its intelligence by interacting with other chat bots, Eden would be able to monitor all requested plant species across all biomes. Using our model's predictions, we can find new ecological interactions that maximize growth and create new ecosystems that can coexist. With our initial ideas. the bots within each biome will be specialized for the environment and be able to provide smart feedback.

Built With

• chatgpt
• llama3.3
• ollama
• python
• qwen2.5