Inspiration
The Winter Olympics is under threat. The number of reliable host locations is currently dwindling as the planet warms. By 2050, out of the 93 potential host sites, only 52 are predicted to still have reliable conditions.
The Winter Olympics is more than just a competition, it's a powerful platform that raises international cooperation and athletic excellence. Frostline was created to raise awareness of this challenge that venues are currently facing by providing data-driven strategies that can help preserve and protect future host sites.
Our primary users are Olympic planning and decision-making committees, who rely on long-term climate viability when selecting these host sites. Additionally, we aim to engage athletes and Olympic fans who care about the future of winter sports. By educating a broader audience, Frostline encourages preventative action and informed decision-making to help safeguard the future of the Winter Olympics.
What it does
Frostline currently aggregates historical climate and environmental data for past Winter Olympic host cities stored in MongoDB. Using the Gemini API, the platform analyzes long-term climate trends to evaluate a city's future suitability and stability as a Winter Olympics host. Frostline also predicts athlete injury risk under increasingly volatile environmental conditions, providing insights that balance logistical feasibility with athlete safety.
In the frontend, the data is displayed in a beautiful map view powered by MapBox, making the user interface simple to use. All the data is displayed when a user clicks on a pinpoint on the map, highlighting the information generated by Gemini API. As well, users are able to compare two locations together, providing a quick glance into which location is the better fit.
How we built it
Frostline was built as a data-driven pipeline that combines historical climate data with generative AI analysis. We collected and normalized climate and environmental data from past Winter Olympic host cities and stored it in MongoDB, using its flexible schema to manage data across different cities, time periods, and climate variables.
MongoDB is used to preprocess and aggregate this data, producing city-level summaries such as long-term temperature trends, snowfall reliability, and climate volatility. These summaries are then passed to the Gemini API, which analyzes multi-year patterns to evaluate future host suitability, climate stability, and athlete injury risk under changing environmental conditions.
By providing Gemini with structured, curated data rather than raw values, we enabled more accurate reasoning and consistent predictions. The results are visualized on the frontend through an interactive Mapbox-powered map, allowing users to easily explore and compare host cities.
Challenges we ran into
One of our main challenges was integrating MongoDB with the Gemini API in a way that allowed the model to effectively use our data. Since Gemini cannot directly query a database, we had to design a pipeline that extracts, aggregates, and formats MongoDB data into structured summaries that Gemini could reason over. Ensuring the data was both accurate and concise enough for AI analysis required multiple iterations and careful prompt design.
Accomplishments that we're proud of
Through building Frostline, we are proud of learning new technology such as MongoDB. We had not utilized MongoDB in the past to store large amount of data and pull it in the frontend. We also gained experience in a good user experience and design, something that is especially important in our platform.
What we learned
Through building Frostline, we learned how powerful generative AI can be when paired with structured historical data. We also gained experience in designing AI systems that support long-term decision-making, rather than short-term predictions, and learned the importance of framing climate data in a way that is accessible to both experts and the general public.
What's next for Frostline
In the future, we plan to expand Frostline's dataset to include future climate projections, economic impact metrics, and sustainability initiatives by host cities. We also aim to refine injury risk modeling and improve visualizations.
Log in or sign up for Devpost to join the conversation.