Entrepreneurs in developing nations have the drive but not the resources to help innovate in their communities. One area of interest is electrical access, with over 1.06 billion people without a reliable access to electricity. One major problem from this is lack of lighting, which makes it near impossible to work at night, with the only affordable replacement being kerosene lamps, which give off toxic fumes. Additionally, they lack an accessible method to reach out to investors willing to help give such resources. All of this was the motivation behind Mulb, an accessible platform for entrepreneurs to bring light to their community, learn smart business practices, and create a sustainable development model.

How we built it

Mulb provides two major services. 1) An innovate way to produce lightning in communities and 2) An accessible platform for entrepreneurs to get in contact with resourceful investors. The first objective is constructed with electrodes, mud, and an LED. The second objective is achieved through the Mulb app. Machine learning algorithms were used to analyze how reliable certain individuals are, NCR will used to catalog and expedite the process of starting and maintaining a business, and Azure (Microsoft Machine Learning Studio) was used to create a method by which two experiments were created to implement usage of machine learning to facilitate lender-lendee connections and trust. The first experiment used a two-class boosted decision tree to create a training model with the 200,000 points in the artificially created dataset to assist the lender in deciding whether a lendee should be trusted when making a deal (and also with what level of confidence). This experiment's training model was able to give the model a 86.2% accuracy when testing. The second experiment also used a two-class boosted decision tree to go through the 600,000 actual data points found off of Kiva (years 2014-2017), and was used to create a training model with 97.3% accuracy to determine what type of payments could be expected from a lendee (monthly vs irregular vs bullet). Together, these two models provide a supplemented, fact-driven by which lenders have a "second-opinion" on their decisions. Xcode was used to provide the layout of the app.

Challenges we ran into

Challenges we faced throughout the hack were reliability of making an electrode cell for the LED, as well as proper integration of the NCR API.

Accomplishments that we're proud of

We're proud of how unique this hack is. We had no plans of doing a hardware hack at the start of this hackathon, but we're happy to have diverged from the typical hackathon-esque project and made something truly outreaching and unique.

What's next for Mulb

Next for Mulb in the foreseeable future is continuation of app development, and finalized product for the sustainable LED light.

Share this project: