It’s time to meet another contender for the Code4Impact Grand Prize – entrance into the HACKcelerator. Meet Owen Leddy, who created his project solo and guided it to success and AngelHack Los Angeles 2016. We chatted with him about his project, the hackathon, and what he’s planning for the future.
Tell us more about the idea behind your project…
My project (under the working title “Zidisha Impact Analytics”) uses data from the microlending platform Zidisha to train a machine learning model that predicts the impact of new microloans currently seeking funds. Microfinance was conceived as a way of alleviating poverty, and there has recently been a growing awareness that it is struggling to achieve a positive impact, hindered in part by high interest rates. Debt from microloans can sometimes worsen a borrower’s situation, or a loan may simply be unhelpful. I wanted to create a tool that would help prospective lenders on Zidisha choose loans that would maximize their positive impact and help the lending platform identify borrowers who might need additional support in order to be able to benefit from a loan. Of all the various online microlending platforms, I chose Zidisha because they display lots of feedback from their borrowers and their interest-free, direct peer-to-peer model genuinely puts impact first.
Was this the original idea, or did you pivot at some point during the hackathon?
I went in with a fairly clear idea of what I wanted to do, and the prototype I ended up with sticks pretty closely to that intention. The method and tools by which I accomplished that end changed somewhat along the way, but I built what I set out to build.
Did you work with a team?
I did not work with a team at AngelHack LA, but the project I worked on is a piece of a larger effort that involves a few of my fellow University of Chicago students. I will be collaborating with them on it moving forward.
Before we get back to the project, tell us about the actual hackathon in Los Angeles. How was it? Tell us your favorite parts!
I had a fantastic time at AngelHack. Even though I did not end up working with a team, I met many brilliant people working on fascinating, exciting ideas and projects. Many of them had helpful suggestions and feedback about my project, which was wonderful. It was an inspiring environment in which to work. The sponsors were also fun to learn about, and I look forward to working with their technologies more in the future.
In your mind, what one thing does every hackathon need to make it a successful event?
Even though sponsors and planning can be important, ultimately I think the people make the event, and the tone everyone collectively sets is the marker of success. As long as people establish a vibrant, collaborative atmosphere where people share ideas, welcome beginners, learn, and fail productively, I think that is the essence of a successful hackathon.
Ok back to your project. We’ve talked about the inspiration, can you tell us a little bit more about the technology? What APIs did you guys use?
Sentiment analysis of comments on Zidisha’s website was very important. The central thesis of the project is that the tone of conversation between a borrower and their lenders is a good proxy for the impact of the loan. An appreciative borrower and happy lenders indicate positive outcome, whereas a struggling borrower and upset lenders indicate a negative outcome. HPE Haven OnDemand’s sentiment analysis API served that purpose perfectly. I used H2O to train the machine learning model – a gaussian generalized linear model – and generate predictions.
What are your future plans with the project?
I am in the process of turning my work from AngelHack LA into a more polished prototype that incorporates more predictive factors and more sophisticated machine learning models and includes a front end where users can see predicted scores for loans currently seeking funds on Zidisha’s website. As I mentioned before, this project is just part of a broader effort. I hope that my collaborators and I will be able to create a set of financial and analytic tools that will help online microfinance platforms maximize positive impact and be financially sustainable.