Noach

Inspiration

In Houston, weather, air quality, and bayou conditions can change quickly. The information exists, but it is scattered across different sources and can be hard to apply to an everyday decision. We wanted to make it easier to answer a practical question: Is this still a good time and place for my activity?

What it does

Noach is a personal environmental planning assistant. Users can save places, plan outdoor activities, and see guidance based on relevant weather, air-quality, and bayou conditions. Its AI assistant can explain those conditions in plain language, with an optional natural-voice response.

Many tools inform users about environmental conditions. Noach connects that information to a user’s actual plans, helping them consider a better time when conditions may affect their health or safety. It also distinguishes forecasts from observed readings and labels uncertainty rather than presenting every signal as a confirmed hazard.

How we built it

We built Noach with Next.js, React, TypeScript, and FastAPI. Gemini helps turn environmental context into understandable guidance, Backboard supports persistent assistant memory, TigerDB stores environmental and user data, and ElevenLabs provides optional speech. We deployed the working web app on Vercel and explored Vultr as a future infrastructure option.

Challenges we ran into

Environmental data comes from different sources, updates at different times, and does not always tell a complete story. We worked through how to handle missing or stale readings, communicate forecast uncertainty, and avoid overstating what bayou observations mean. Integrating voice and deploying the full experience also took iteration. We explored Vultr GPU hosting, but GPU availability prevented us from using it for this release.

Accomplishments that we’re proud of

We brought environmental data, activity planning, AI guidance, remembered preferences, and optional voice into one working experience. We’re especially proud that Noach does more than display numbers: it helps people use those numbers to make decisions about their own places and plans.

What we learned

Useful AI depends on trustworthy context. A fluent answer is not enough if the underlying data is stale, incomplete, or presented with too much confidence. Building Noach taught us to treat data freshness, clear labeling, and personal relevance as core parts of the experience.

What’s next for Noach

We want to improve personalized recommendations and proactive alerts, expand voice interactions, and add more verified local data. We also plan to revisit Vultr infrastructure as we grow the system beyond this hackathon release.