100 million days in Vienna: A story of APIs & AI in tourism.

From 22 APIs to one perfect day. Discover the AI architecture behind Vienna's personalized travel planner.

#1about 3 minutes

Introducing the "Make My Day" tourism prototype

The application's tech stack is introduced alongside key vocabulary from the tourism industry to set the context for the project.

#2about 9 minutes

A live demonstration of the itinerary planning app

A walkthrough of the "Make My Day" web application shows how a user answers a quiz to generate a personalized, actionable daily plan for Vienna.

#3about 5 minutes

Understanding the system architecture with the C4 model

The C4 model is used to explain the high-level architecture, showing how the application interacts with the central data hub and various external services.

#4about 7 minutes

Designing APIs for data aggregation and UI plugins

An exploration of specific APIs, including an opinionated location aggregation service and an oEmbed-based service for embedding UI components like maps.

#5about 6 minutes

How the "dress a skeleton" planning algorithm works

The core logic for creating itineraries is a 10-step procedural algorithm that incrementally fills a pre-defined day structure with matching points of interest.

#6about 4 minutes

Using machine learning to classify touristic data

Logistic regression is used to classify places of interest based on atmosphere and weather suitability, highlighting the challenges of human bias in data labeling.

#7about 4 minutes

Building a data aggregation and enrichment pipeline

An ETL-like pipeline retrieves structured and unstructured data, uses NLP to link it to points of interest, and aggregates it into a knowledge graph using schema.org.

#8about 4 minutes

Key learnings and evolving the prototype into a framework

The project's key takeaways, including the limitations of NLP in Node.js, led to evolving the prototype into a reusable framework for other tourism destinations.