For those familiar with hackathons, this was anything but. 7 hours, 4 people, 1 team, 0 ‘winners’. But what we gained was invaluable insight into the platform and its potential — plus next steps into user validation.
A neuroscientist, UX designer, mathematician, and programmer walk into the dedicated Make Space, courtesy of Offerzen’s awesome Make Team: Ben, Helene, and Stephen.
The day started like most of my days: coffee, computer, and confusion: with so many possibilities, what shall we focus on!
We knew from the start that the day was not meant to produce a functioning product. We wanted freedom to explore strange ideas and fantastical possibilities.
The official Make Days have a similar atmosphere, which I have been fortunate enough to experience first hand — as a hacker and as a “Make Master”.
The difference here is the team.
While the usual suspects for the organised ‘hackathons’ are programmers with 4+ years experience, I knew I wanted insight from alternative backgrounds and contexts.
It was also clear that a recent update to the Root Insurance API meant the existing Python module needed to be updated (credit to the initial work done by Brandan Ball and Christo Goosen). I’ll be sending a pull request to the root-community version when I think it’s ready.
root-insurance-python - python port of https://github.com/RootBank/root-insurance-ruby
We spun some ideas together both on the Insurance and Banking sides of Root, including ideas for
- Automated data insurance (in case of a security breach)
- Providing IP insurance [for code] through Github/Bitbucket webhooks
- Servicing a number of savings ‘tricks’ through a central app that can be opted in (e.g. rounding to the nearest ‘X’, donating to charity when buying tea instead of coffee [which is still < coffee], creating ‘invisible’ savings ‘pots’ in the RootCode, a ‘panic’ mode for your bank account to minimise expenditure, etc.)
- Providing a recipe book for defining and activating features akin to IFTTT, without needing programming experience
Ultimately, we decided that a core part was missing: knowing what users really wanted. The solution?
An AI/NLP research assistant that asks users questions through a chat interface (FB messenger, Google Assistant, Amazon Echo, etc.), built using Dialogflow. This is additionally interesting as it is the reverse of how chatbots are normally designed: here the chatbot asks the user a question, to which they respond.
While we are not ready to showcase the artificial research assistant just yet, suffice to say there is additional AI being incorporated in the analysis of responses and content of questions.
For more artificial assistant goodness check out my recent talk at the Deep Learning Indaba X Western Cape, helping strengthen Machine Learning in Africa! And for more talks from the event, visit the author’s channel!
Ok, so what did we achieve?
Exactly what we set out to do:
explore, discover, refine, iterate
We learnt that writing good code is tough, but building a good product requires user feedback regardless how good you think it is.
In the spirit of the Make Day hackathons, we created a chatbot on Dialogflow, we hacked on the Root [Insurance] API, and had fun doing it!
I look forward seeing you at future Make Days.
New projects for the days are being developed right now, including a computer vision idea [hopefully]…