"On the road" — a journey companion
Your AI-Powered Adaptive Travel Assistant
Plan less. Experience more.
Yolda turns a natural conversation into a fully-optimised itinerary, then silently rebuilds it in real time whenever your trip veers off track — so you stay present in the journey.
The Challenge
Modern travel planning is still fragmented, stressful, and difficult to manage in real time. Every traveller faces the same four frustrations.
Trips quickly go off track. A closed venue, an unexpected detour, or a simple change of heart unravels the whole day with no recovery path.
Most apps cannot adapt live. Once a plan is generated it stays frozen — users are left replanning manually under pressure.
Too many choices make planning difficult. Sifting through dozens of reviews, blogs, and maps to build a coherent route wastes hours before the trip even starts.
Building a good route takes effort. Coordinating opening hours, travel distances, budget limits, and personal interests into a single optimised schedule is a complex task.
Core Innovation
A "Live" agent continuously creates, monitors, and repairs your itinerary — combining an agentic AI framework with a constraint-solving optimiser.
You describe your trip in plain language. The agentic framework identifies your intent, extracts constraints, and hands off to the right agent.
The Planning Agent selects the best POIs. The CP-SAT solver then sequences them into the mathematically optimal daily route.
GPS tracking detects deviation. The rerouting service instantly re-runs the solver on remaining POIs from your current location — no manual action needed.
Autonomous Agents
Three specialised agents power Yolda's intelligence. A LangGraph orchestrator assigns every user request to the right agent automatically.
Transforms a conversational description of a trip into a fully structured day-by-day itinerary. Powered by Gemini LLM via Google Vertex AI, it understands vague intent, extracts constraints, and selects the best candidate POIs before handing off to the CP-SAT solver for sequencing.
Handles in-trip modification requests such as swapping, removing, or adding POIs. When the user says "skip the museum, find something nearby instead," the Modify Agent interprets the change, updates the plan state, and re-runs the optimiser — all without losing the overall trip structure.
Answers questions about any point of interest in the user's itinerary — history, opening rules, tips, nearby alternatives — using AI-generated contextual knowledge. Critically, it operates in read-only mode: it provides information without mutating the active plan.
Optimisation Engine
Google OR-Tools CP-SAT finds the mathematically optimal POI sequence within hard real-world constraints, maximising a blended score that balances user preferences and ratings.
Features
From AI-powered conversation to real-time rerouting — every layer is designed to keep you moving, effortlessly.
Describe your trip in plain language. Yolda's LangGraph multi-agent system backed by Gemini LLM extracts your intent, budget, pace, and preferences — no forms, no filters.
OR-Tools CP-SAT sequences every POI within your opening-hours window, budget cap, and daily activity count — producing the highest-scoring achievable itinerary.
Real-time location updates are stored in Redis Geo sets. The tracking service continuously monitors your proximity to the planned route and triggers alerts on deviation.
On deviation, the rerouting service re-runs CP-SAT on remaining POIs from your current GPS position and rebuilds the itinerary instantly — no manual action needed.
Under the Hood
A modern, production-grade stack chosen for reliability, scalability, and developer velocity.


Product Demo
Watch a full walkthrough of Yolda's core features — from conversational planning to live rerouting on a real trip.
The People Behind It
Six engineers who designed, built, and shipped Yolda end-to-end — METU Computer Engineering, 2026.
Get in Touch
Have questions about Yolda or want to learn more about the project?