|About the Job|
August 2012 – February 2013
Participated in the development of 3 products of this Artificial Intelligence company.
Foris Predict: Early detection of student desertion
The Desertion module at Foris is a software that identifies students that are at risk of dropping out from college with hopes of taking actions to prevent this. It achieves this by applying complex data analysis and machine learning over the students history. The objective is to detect this potential drop outs early on so that the University can take actions to prevent them from leaving. I worked on this project full-time at the beginning of 2013.
Foris is a company that was founded in 2010 by professionals specialized in Analysis and Software Development that provides solutions based on Artificial Intelligence. Their solutions are mostly oriented to the higher education industry, with two very successful products that have bumped them to the position where they stand today: DarwinEd and Deserción. The company’s headquarters are in Santiago, Chile, and it serves over 40 Universities spread across 8 countries.
Foris Predict embodies a methodology that analyzes and helps comprehend the desertion patterns that get presented in higher education establishments in order to increase the rates at which students graduate.
Using a profound analysis of the student data when they sign up and through their performance in class, Foris Predict allows universities to:
- Estimate the probability of individual desertion for each student.
- Understand the causes why a student would desert.
- Perform a following of the individual management of for the retention of each student in risk.
Using Data Mining techniques, all the historic data from the educational institution gets explored, as well as the demographics of the analyzed students, which then yields the specific patterns of desertion and therefore the predictive models. At the same time an analysis of the root causes of desertion allows for categorizing the deserters by type and coordinate and prioritize assistance procedures for each student group.
There were several versions of this Software, all based of the same prediction model. The company would set out to create a specific solution for each of their clients fulfilling each of their specific requests. The version that I got to work on was one of the first pieces of software developed for this product, it was a Java App in which users could either input manually the data required by the prediction module and get the results on screen or select a database file with the info of all the students and generate an Excel file with the resulting predictions.
On my first day at Foris my manager said to me: “Do you know how to do this?”, I replied: “No”, to which he added: “You have until tomorrow to get it done.” It caused quite the impression on me. I didn’t think I was going to be able to make it, so I felt somewhat upset at such an imperative request. It still surprises me to this day that I was indeed able to finish it by the next day and it taught me a valuable lesson on how a good manager can bring out the best out of you.
Foris Predict was a great project to have worked on. I got to work closely with a brilliant mathematician that deciphered the model to achieve the best predictions, a model that he came up with after months of researching university data. The resulting product became one of the most successful projects within Foris that they proudly market and with which they serve their international clients to this day.
Through the development of the project I acquired valuable experience regarding machine learning, prediction models, data mining and analysis, statistics, and a lot of teamwork.
- Development of Smartphone Applications.
- Development of Analytics Tools.
- First time working in a team under the leadership of a Project Manager.
Frameworks, Software and Technology used
- Apache POI
- Eclipse IDE
- Subversion (SVN)