trAAvl — AI Copilot for Disrupted Travel

Inspiration

We’ve all lived the same chaos: delays with little guidance, gate changes mid-walk, and long lines at customer service while trying to rebook, find a hotel, and figure out baggage. We built trAAvl to give passengers a single place to understand what’s happening and what to do next—while also giving ops teams a clearer way to triage disruption impact.

What it does

Passenger Portal

AI Travel Assistant: Answers questions about your trip, airport logistics, aircraft type, and disruption steps (with guardrails to avoid making up details).
Real-time Flight View: Flight status, delay indicators, and gate-change notifications (live where possible, simulated for demo scenarios).
Delay Risk Insights: Delay likelihood based on weather context + historical-style heuristics (explainable, not “black box”).
Connections Coach: Estimates connection feasibility using terminal/gate walking-time models and shows step-by-step transfer guidance.
Disruption Recovery Hub: Rebooking guidance, nearby hotel suggestions, and ground transport options (rideshare/taxi pickup instructions).
Baggage Help: Quick path to report mishandled baggage and track status updates (demo workflow).

Employee Portal

At-Risk Connection Dashboard: Identifies flights with high misconnect risk and passengers needing assistance.
Assistance Queue: Tracks wheelchair/cart/escort requests with dispatch/resolve status.
Scenario Simulator: Triggers delay, gate change, hold, cancellation, parking/rideshare updates to demonstrate end-to-end impact in real time.
(Optional demo module) Volunteer/Overbooking flow: lets passengers submit a compensation bid and simulates acceptance logic.

How we built it

Frontend: Modern React UI with a consistent “glass” design system, motion, and reusable components for both passenger + employee experiences.
Backend: Secure APIs + realtime updates to push events instantly (gate changes, holds, alerts) to the right screens.
Database: Structured trip + segment records plus an event stream model for updates (delay, gate change, baggage status, etc.).
AI: A constrained passenger assistant designed to use known trip context + approved sources, and to say “I don’t know” when data isn’t available.
Integrations: Flight status (third-party), weather (forecast API), hotels (API or demo dataset), maps/gate routing (airport graph model for DFW/CLT).

Challenges we ran into

Realtime reliability: Making sure every event (gate change, delay, hold) instantly updates both passenger and ops views without refresh.
Timezone correctness: Keeping flight times consistent across airports and displaying them correctly per location.
Safe AI behavior: Avoiding hallucinations by restricting the assistant to verified trip context and web-sourced info only when requested.
Consistent UI/UX: Preventing one-off styling by enforcing design tokens and shared component patterns.

Accomplishments we’re proud of

End-to-end demo flow that feels real: delay → gate change → connection risk → hold event → assistance request → recovery options
Passenger experience that reduces uncertainty with a single timeline and clear next steps
Ops experience that prioritizes who needs help first instead of reacting blindly
A practical “buildable” scope: real APIs where feasible + simulator where airline-internal data isn’t available

What we learned

Airline problems are rarely “one screen” problems—solving them requires coordination + realtime + clear ownership
Explainable predictions beat fancy models in a high-stress UX
A good event system (and role-based access) is the difference between a demo and something believable