Inspiration

On days when storms roll in, your calendar shouldn’t force you to gamble between your plans and your safety. We’ve all had that moment where the sky looks terrible, but the meeting feels non‑negotiable, and safety loses to obligation. Generic weather warnings freeze hundreds of millions of person‑hours for threats that often never hit home. In Oklahoma alone, 152 tornadoes in 2024 turned “weather interruptions” into a normal part of life, and in those critical 15‑minute windows people are alerted they often burn precious time doom‑scrolling just to figure out if the danger is real for them. Our AI‑powered weather + calendar app steps in right there, we make “doing the safe thing” the default, so you don’t have to choose between “maybe I’ll be okay” and your life.

What it does

1. Plan: The Hazard‑Aware Calendar Dynamic Risk Analysis We turn a static calendar into a hazard‑aware planner by adding an AI Safety Analysis layer to every event. In the Context‑Aware view, the app flags high‑risk time windows using location‑specific NWS data, so a 3 p.m. meeting is no longer “just a meeting” if it sits inside a tornado watch.

One‑Tap Mitigation From any event, users can shift to safety in a single tap: “Update to virtual” or “Reschedule meeting” directly from the calendar view. This removes the psychological friction of canceling by automating the professional fallout. Invites are updated, messages are sent, and you stay safe without looking unreliable.

2. Build: Spatial Safety Maps Vertical Safety Routing Our Safety Radar does not just show traffic. It visualizes storm progression, flood‑prone segments, and route exposure over time. Routes are labeled in plain language (for example, “Home → Office: SAFE for next 45 minutes”) so users instantly see where the risk is, not just where the road is.

Predictive Navigation We intersect your live GPS path with NWS warning polygons to compute safe travel windows before you leave. The engine prioritizes vertical safety by favoring higher‑elevation segments and steering drivers away from flash‑flood zones, reimagining navigation around survivability instead of speed.

3. Maintain: The Guardian AI Advisor Agentic Support The Guardian AI chatbot acts like a proactive safety officer living inside your calendar and map. Instead of generic alerts, it gives clear, situation‑specific guidance: “I recommend canceling outdoor transitions after 4:15 p.m.; move your car to covered parking on higher ground.”

Actionable Checklists When conditions escalate, the app generates context‑aware preparedness lists tailored to your reality: an emergency car kit before a blizzard commute, a go‑bag checklist when severe storms line up with your evening plans. Users are not just warned. They are walked step by step toward a higher baseline of readiness.

4. Persist: Resilient Infrastructure Safety‑Integrated Dashboard The daily timeline becomes a safety‑aware control center, tying together your commute, classes, meetings, and gym sessions with live weather impact tags. High‑risk blocks are highlighted directly on your schedule so you can rearrange your day while you still have options, not when the sirens start.

Offline‑First Reliability Because infrastructure fails when you need it most, critical safety data is cached ahead of time. Elevated evacuation routes, building refuge maps, and key checklists remain available even if cell towers or data service go down. The app is designed to keep guiding you when the rest of your tools go dark.

How we built it

Gemini + Google Calendar: We used the Gemini API to read NWS data alongside Google Calendar events, turning raw alerts into clear survival guidance and auto‑drafted reschedule emails so acting on warnings feels natural, not awkward.

Google Maps + Safety Routing: Using the Google Maps Essentials tier, we styled maps for an emergency mode and built elevation‑aware routing that prefers higher ground to avoid Norman flash‑flood zones during severe storms.

Data + Offline Resilience: Real‑time NWS polygons are intersected with GPS paths and stored with campus safety GeoJSON in MongoDB Atlas, while a local‑first design lets users download safe routes and plans for full offline use during outages. ​ Frontend + Auth: Next.js and Tailwind power a state‑aware UI that shifts from green to red with threat level, and Google Sign‑In keeps “safe circles” plus home and work locations synced securely across devices.

Challenges we ran into

OAuth with Expo Go Google Sign‑In repeatedly failed with redirect_uri_mismatch and policy errors because Expo Go uses special redirect URIs. We had to carefully align the Expo proxy URLs with our Google Cloud Console settings and adjust the auth flow to make login reliable.

Tokens and Calendar sync Implementing secure refresh tokens plus Google Calendar’s incremental syncToken flow turned out to be nontrivial. We had to coordinate backend logic to refresh tokens safely, handle invalid or expired syncTokens, and gracefully fall back to a full resync when Google rejected a delta request.

Accomplishments that we're proud of

We are most proud that we built something that actually feels like a calm guide when the weather turns dangerous. Our system watches official weather alerts and your movement so it can give you a simple, personal sense of how much time you have and where your safer options are, even if the power or cell service goes out. At the same time, we know it is emotionally hard to step back from work, school, or plans you care about, so our AI is designed to feel more like a supportive friend than a siren: it connects the weather to your actual day, suggests better options, and even helps handle the awkward rescheduling so that choosing safety feels like protecting your future, not abandoning your responsibilities.

What we learned

We learned that building a “safety app” is really about understanding human behavior, not just stacking APIs. As a team, we had to balance ambitious ideas with a ruthless focus on what actually helps someone make one better decision on a stressful day, and that meant simplifying flows, communicating risk in plain language, and leaning on each other’s strengths when the tech (or the weather data) fought back.

What's next for ArkOut

Institutional API Integration: Our next phase involves integrating directly with university alert systems (like OUDaily or campus police) to ensure our "Social Proof" notifications are perfectly synced with official emergency protocols.

Low-Bandwidth Optimization: We aim to further optimize our offline survival mode, ensuring that high-resolution map tiles and AI escape plans can be stored using minimal device storage for students with limited phone capacity.

Share this project:

Updates