Predict

Fodé Diop — Thu, 30 Apr 2020 22:59:23 +0000

Predict

Category Submission: Exciting X-Factors

Predict is a simple but powerful Machine Learning SMS API that can predict the gender of an individual from their first name.

Demo Link

https://twilio-predict.herokuapp.com/redoc

Link to Code

https://github.com/diop/predict

Example

Web --> https://twilio-predict.herokuapp.com/predict/michelle
SMS --> https://twilio-predict.herokuapp.com/hook

Software Requirement

Python 3.7 or later
FastAPI
scikit-learn
Jupyter Notebook
A Twilio account - sign up

API Documentation

How I built it (what's the stack? did I run into issues or discover something new along the way?)

As I mentioned in my previous post, SMS is a great equalizer when it comes to democratizing access to artificial intelligence. My idea was to build a Machine Learning model and be able to serve it via SMS as a proof of concept.
With SMS one does not need an internet connection or powerful mobile phones to have access to life-changing services. Only a cellular connection will suffice. What if we can leverage that basic text message infrastructure to empower the people who need it the most, and give them access to intelligence augmented services? Because then all of a sudden a farmer in a remote area can text a photo of his crop to get diagnosis or prediction on the crop's health. Somewhere else, in a remote area, a nurse can text the photo of skin leisure and get diagnostic on that particular disease. The possibilities are endless!

In this hackathon, I wanted to focus on the problem-solving aspects of SMS because most people who live in the developed world have no idea what it's like to live without adequate internet access.

There were 3 parts to my process. The first step was to gather the dataset of names, male and female. The second step was to do some data analysis and make sure that the dataset did not have any missing or duplicated values. I also needed to check if the dataset was balanced as well, meaning that I had a good even ratio of female names to males ones, not to introduce some biases in the machine learning model. Finally, the third step was to build the model and deploy it so that the Twilio client could consume it.

This process is almost typical for any machine learning model that needs to be deployed to production. Let's help make the world a better place!

Democratizing Access to Artificial Intelligence

Fodé Diop — Fri, 24 Apr 2020 01:14:12 +0000

Hello, Twilio hackers! I'm a little late to the party, but there's still a week left in the hackathon. So why not put a submission in?

While I was traveling to visit my family in Sénégal, West Africa, the whole Covid-19 mess happened. Instead of sitting idle and not do anything, I decided to do a Machine Learning bootcamp at Vivadata Dakar cohort and finally add some artificial intelligence to my arsenal. Boy, I'm glad I did!

Professor Andrew Ng was famously quoted for saying "Artificial Intelligence is the new electricity.". It's going to affect and improve every aspect of our lives. But access to this nascent and game-changing technology is fairly limited. It's still very specialized knowledge. What can we do to get it into the hands of the people who need it the most, like here in West Africa?

SMS is a great equalizer. You don't need the internet (only cellular connection is enough), and you certainly do not need a fancy super powerful phone. The idea is to build a simple enough API endpoint to serve Machine Learning models, then leverage the Twilio system to allow for any participants to get results back by sending messages to a specified Twilio phone number. I've seen some seemingly working prototypes out there, but they work like black boxes. We don't really know how they work.

My goal with this hackathon is to utilize the Machine Learning knowledge I've recently acquired, to build an intelligent SMS system prototype from scratch and to open source it. Since there isn't much time left, let's get to work.

Happy hacking!

Poster Photo by Trevor Cole on Unsplash

Forem: Fodé Diop

Predict

Predict

Category Submission: Exciting X-Factors

Demo Link

Link to Code

Example

Software Requirement

API Documentation

How I built it (what's the stack? did I run into issues or discover something new along the way?)

Democratizing Access to Artificial Intelligence