Inspiration
As immigrants and travelers (Pedro, Andres, and Salvador), we created DestinAI because we experienced the pain of being in new places but never knowing what to do or where to go. We recognized that existing platforms like Expedia give generic results, while LLMs like ChatGPT hallucinate, creating unreliable plans with non-existent hotels and outdated prices, leaving the work of piecing everything together to the user.
What it does
DestinAI solves the problem of unreliable trip planning by using a "divide and conquer" strategy to deliver a single, cohesive, day-by-day travel plan that is tailored, optimized, and ready for you. Our system dramatically reduces the risk of hallucination by delegating tasks. The AI analyzes 45 distinct user inputs—from your age for legal checks to your budget and 36 specific interests—to become a reliable alternative to broad AI prompts.
How we built it
We implemented Google's Agent Development Kit (ADK) to build 5 specialized AI agents: Transportation, Flights, Events, Hotels, and Weather. These agents communicate seamlessly using the Agent2Agent (A2A) protocol to process the 45 inputs and delegate tasks between specialists, ensuring a reliable, singular itinerary as the output.
To make the AI accessible, we also developed a cool, interactive web page for the user interface, utilizing standard frontend technologies: HTML, CSS, and JavaScript. This front end gathers the 45 user inputs and displays the final, cohesive travel guide generated by our collaborative agent system.
Challenges we ran into
The primary challenges were technical, specifically implementing and stabilizing the A2A protocol. We also faced a significant organizational challenge in finding good teammates, which meant we started development a day late, and after 1 of them leaving we had to rebuild our workflow quickly.
Accomplishments that we're proud of
We are proud of building a functional, reliable multi-agent system using ADK and A2A. We successfully created an AI that can analyze 45 distinct user inputs and return a cohesive, highly tailored, optimized travel plan.
What we learned
We learned valuable lessons in new technologies (ADK/A2A), the importance of strong teamwork, and the need to never give up despite early team and technical setbacks.
What's next for DestinAI
Our next steps are to better the user interface and to work with other cities, as the platform currently only operates for Miami. Also we want to expand the events that the application can look for, seeing that right now it is limited.
Log in or sign up for Devpost to join the conversation.