Inspiration

As ski lovers, we’ve always found that planning a trip takes away from the fun. Choosing the right resort, finding flights, booking hotels — it’s a maze of tabs, dates, and decisions. We wanted a way to bring the joy back into planning — something that feels as smooth as gliding down fresh snow. That’s how SlopeMate was born: an AI-powered companion that plans your entire ski trip effortlessly.


What It Does

SlopeMate is a multi-agent AI system that collaborates to design your perfect ski trip. It understands your preferences, asks clarifying questions, and coordinates multiple agents to generate a complete itinerary — from flights to resorts to hotels.

  • Orchestration Agent — Manages conversations and integrates all results into one seamless plan. Ask clarification question if any required information is missed.
  • Resort Agent — Finds ski resorts matching your style and preference by prioritizing to use API info. Fallback to use LLM knowledge if no info is found via API.
  • Hotel Agent — Recommends comfortable and convenient stays by prioritizing to use API info. Fallback to use LLM knowledge if no info is found via API.
  • Flight Agent — Locates the best flight options for your travel dates by prioritizing to use API info. Fallback to use LLM knowledge if no info is found via API.

How We Built It

We built SlopeMate using a multi-agent architecture powered by large language models. Each agent operates semi-independently with its own prompt specialization and external API integration. The Orchestration Agent acts as the control center — it interprets user input, delegates tasks to the right agents, and merges results into a natural conversation flow. We combined:

  • Chat-based user interface
  • LLM-based reasoning and dialogue management
  • API calls for real-world resort, hotel, and flight data
  • Context synchronization between agents for collaborative decision-making


Challenges We Ran Into

  • Agent Coordination: Getting agents to share context efficiently without flooding each other with data.
  • Orchestration Logic: Teaching the OrchestrationAgent when to clarify vs. when to decide.
  • Response Consistency: Ensuring results from different APIs feel unified and coherent.
  • User Experience: Making interactions natural and human-like, not robotic or repetitive.
  • In-memory Cache: One challenge we faced was that the agent often lost the context of the user’s previous answers. To solve this, we built an in-memory cache that stores key user information and carries it across different conversation states. This cached data ensures that sub-agents have the required inputs to function properly. If any information is missing, the Orchestrator agent will step in and ask follow-up questions to fill the gaps.

Accomplishments That We're Proud Of

Built a fully functional multi-agent collaboration system that feels conversational and intuitive. Designed an orchestration flow capable of dynamically managing specialized AI agents. Created an experience where users can plan a complete ski trip through a single, friendly chat. Proved that multi-agent AI can meaningfully improve real-world trip planning experiences.


What We Learned

We learned how complex agent communication can be — and how rewarding it is when done right. We discovered the importance of prompt engineering, context control, and dynamic reasoning for multi-agent orchestration. Most importantly, we saw how AI systems can feel collaborative, not just reactive.


What's Next for SlopeMate

We plan to let users provide external Instagram or TikTok video links as tentative locations when they’re unsure of the exact place.

Looking ahead, we aim to expand SlopeMate beyond ski trips — to beach vacations, hiking adventures, and city getaways. We’ll keep enhancing agent reasoning, real-time data integration, and personalization. Our ultimate goal is to make AI trip planning effortless, enjoyable, and perfectly tailored to every traveler — in every season.

Built With

Share this project:

Updates