We all have experienced the stress of finding the perfect trip. Just a a few days ago I tried booking a trip some ski-resort in Europe. The process is straightforward, but awfull. First you have to know which resort you are going to, then you need to figure out a suitable and cheap airport. But flight prices vary a lot, and if you find a cheap flight for one week, it doesn't help unless there is also good hotels available for the same date.
We would need something that keeps monitoring the internet for good deals, telling us when something good pops up. It should work like a human, understanding requests like "I want to ski next winter", as well as user preferences like "I like french food" or "I wont sleep in a dormitory" or even things like "I like pasta".
The internet is bustling with all kinds of travel analysers, but they are focused on users who know what they want, and they don't offer humanlike interfaces, instead resorting to aggressively pushing discounts and "good deals".
Because trip finding is such a chore, we tend to prefer longer trips because of the chore/pleasure balance. This is not ideal for flight operators because longer trips = more costs, but the flight cost remains constant. Using Aino you can easily find good/cheap/perfect short trips, for example to go hiking with your friend for a weekend.
What it does
Agent Aino is your companion travel agent. It finds good trips for you based on your preferences, blogs you have linked to it, images you send it, and your answers to its questions.
Agent Aino allows you to maintain a trip search by yourself, with you girlfirend, your friend in Chile or your entire school. You can use it to find a deal that synchronizes your flights from any number of locations. It can support any level of general requests, like "I want to fly to a place that serves pasta".
How we built it
Python, Natural language processing (nltk) and ML based on Heroku using AWS Rekognition for image recognition. After ideation we divided the challenge between our teammembers, constantly pushing to our agile git.
Challenges we ran into
NLP and ML is very difficult, correctly analysing unstructured data like blogposts is a very contemporary challenge. We also had tech challenges with python, which is not anyones primary language in our team.
Accomplishments that we're proud of
The working demo is something we are extremely proud of. It analyses any website or image, and return an optimal trip choice based on keywords and flight data.
What we learned
NLP&ML crawling, python teamwork, having fun, hackathons, life, big teams, git, geocoding
What's next for Agent Aino