Inspiration

We've all tried to arrive somewhere, and it wasn't quite what you expected. And moreover, some things may not have been obvious when booking and appear different at arrival. These can vary from minor inconvinences to major deal-breakers. Especially for people with mental issues, hidden disabilities or hyperfixations. General awareness and availability for these groups are, fortunately, becoming more and more normal, but we've not yet seen it in the booking industry. How are these people supposed to find a place that offers individually tailored filters and setttings? This is the gap we are filling with this project.

What it does

If you want to book a stay somewhere, you look at a few reviews on the booking site, what offers the provider has - activities, bars, elevators, nearby transport etc. and maybe a third-party review. But this require you to go through many different sites, source,s reviews and still rarely conver the full picture. This project takes knowledge, data and information from all these sources and offer a one-stop for both the average booker, but moreso people with special needs.

Is there braille support in the hotel?
Has the hotel previously shown awareness, or lack thereof, for people with special needs?
Does the hotel have a sensory-sensitive friendly environment, e.g. bright lights, loud noises, or strong smells, which can cause discomfort or distress.
Are people complaining a lot about noises from roadwork recently?
Is it possible to do tailored routines for people who struggle with the disruption of their daily routines?
Is the hotel adequate for service animals? All these are maybe irrelevant for most of us, but for a select group, these are essential for finding the right place.

How we built it

A series of data sources that leads into agents that summarize data and gives the prioritizes the users special needs.

Challenges we ran into

We spent about 45 minutes trying to get Claude to work, before bing notified that it is not supported by the event, even though it looks ready in the model serving.
Problems with the quality of the Llama model and several data hallucinations.