EmberAI

Homepage
Risk forecast sample result
Risk forecast sample result
Risk forecast page
About page

Inspiration

From coast to coast, wildfires have increasingly become a threat not just to the directly affected areas but to neighboring regions as well. Witnessing how the smoke from Canada’s 2023 wildfires traveled hundreds of miles to blanket New York in haze was a stark reminder that these disasters don’t recognize borders. Meanwhile, recurring fires in places like Los Angeles further drive home the point that wildfires can ignite almost anywhere at any time. Knowing that the United States sees an average of 68,707 wildfires every year (from 2001 to 2020), we felt compelled to act. We wanted to build something that could both educate and empower. Our goal was to create a tool accessible to everyone—helping communities understand their local wildfire risk and take proactive measures to protect their homes and loved ones. That vision fueled our development of a dynamic machine learning model and a user-friendly website. It’s our hope that by making risk assessments more transparent and widely available, we can play a part in safeguarding neighborhoods from the growing menace of wildfires.

What it does

It provides a weather forecast for a selected area and predicts a value from our wildfire risk index. The risk index categorizes the level of wildfire risk as low, medium, or high, based on the estimated acreage of land that the model identifies as being at risk of developing a wildfire.

How we built it

Using the collaborative feature extension on VS.code, we built a Frontend using html, css, and javascript, a backend featuring Flask, and a multi-regression classification machine learning model using the NumPy and Pandas library and tools like Sklearn in python. Our backend consisted of multiple linkages with not just parts of our UI but also with real-time weather data APIs, our machine learning model, and online databases. It was a tedious but rather worthwhile process to build the process for our model. Lastly, the backend consisted mainly of Flask integrations and the use of APIs, and the Herbie python library (for more weather data) to conduct search and retrieve calls for real-time weather data, our local files, and our machine learning model.

Challenges we ran into

Part of the process was learning on the job. Two of our previous members previous novices and had little knowledge of full-stack development and machine learning in general. We watched youtube tutorials, found explanations for errors on stack overflow, and finally created our own implementations of the machine learning model from scratch so it tailors to our specific purposes. In regards to obtaining our training data, we were very picky about it. Due to the lack of datasets that fit our need, we ended up dynamically creating our own dataset.

Accomplishments that we're proud of

We’re incredibly grateful for the opportunity to showcase what we’ve learned and built during this hackathon journey. It’s been an exhilarating experience diving into new concepts, experimenting with machine learning, and pushing the limits of what we thought we could achieve. Through collaboration and perseverance, we developed an ML model that we’re proud of—not just for its functionality, but for the countless hours of brainstorming, debugging, and refining that made it possible. This project reflects not only the technical skills we’ve gained but also the invaluable lessons about teamwork, creativity, and resilience. We’re thankful for the mentors, peers, and inspiring projects around us that fueled our growth and helped transform our ideas into something tangible and impactful. Imma be honest, Ember AI? That was pretty fire. Even if it may not be the winning project, to us (and hopefully others), it was an opportunity to inspire and help grow our mindsets and skills.

Ps. Team AarWil Ray (pronounced Air-Will- Ray) Aaron: “I wanted to make jumping buttons, but the scroll animation isn’t bad” Wilbur: “The fire is fire, I love my UI designs” Raymond: “data curation is some much fun. But I should def write a program to do that next time.”

What we learned

Machine learning is often portrayed as intimidating and perceived by people as very difficult. However, in this hackathon, we found out it’s far more approachable than it seems, especially with the right tools and mindset. Leveraging Python, Flask, APIs, and libraries like Scikit-learn, Pandas, and NumPy, we can quickly learn to build and deploy a practical machine learning model (Our minimum viable product (MVP- yes glasses guy, thanks teaching us the jargon)). Like for example, just last Friday, we wouldn't even be able to fathom how the ml-based and API integrated (to StockFish) chess glasses would work. Now, we've at least understood the gist: computer vision and machine learning recognition through already available and highly accurate datasets (which we didn't have so we did by hand), APIs to stockfish AI, and a feedback system via output of stockfish chess coordinate predictions into chatGPT's audio convertor function and then direct integration into the hardware speaker for audio. Concepts? Understood. Technicalities and execution? Exciting (W project btw)! Learning to compartmentalize workloads, we fostered essential project management skills and placed our experts and beginners in the areas they best shone. Hackathons, in particular, are an incredible way to bring this learning to life. These events combine collaboration, friendly competition, and peer motivation, all the while, still offering the opportunity to draw inspiration from innovative projects- amazing. They gather diverse and talented groups of individuals who share a passion for coding, AI, and engineering, creating an environment that cultivates creativity, problem-solving, and real-world application. For educators, promoting these experiences is a chance to ignite curiosity and equip students with the tools and confidence they need to thrive in the tech-driven world. For us students, it is an opportunity to explore cutting-edge technology, develop practical skills, and connect with a community of passionate peers who inspire and challenge us to grow.

What's next for AarWil Ray

Team AarWil Ray is dedicated to continually attending and competing in hackathons, not only to challenge ourselves but to embrace every opportunity to learn and grow. While winning would undoubtedly be a rewarding achievement, our primary focus is on the journey itself—gaining new insights, sharing ideas, and improving our skills. Each hackathon presents a unique chance to collaborate with others, draw inspiration from their creativity, and learn from both their successes and our own setbacks. By approaching each competition with a mindset of curiosity and resilience, we aim to make the most of these experiences, fostering personal and team growth while celebrating the joy of innovation and discovery.