Aaron Zampaglione

About

I'm Aaron Zampaglione. I'm your typical nerd who is curious about anything related to technology: machine learning, statistical analysis, computational biology, cloud computing, web apps, open source, and "hacking" anything I don't understand. Outside of tech, I'm interested in the hard sciences, body building, nutrition, and literature relating to the human condition. You can usually find me behind my screen, behind a book, at the gym, or at the beach thinking of ways to improve the world around me.

"... Your daily life is your temple and your religion. Whenever you enter into it take with you your all...”

- Kahlil Gibran

Education

Harvard University

Master of Science
Data Science

Florida Institute of Technology

Bachelor of Science
Computer Science

Budapest Unversity of Technology and Economics

Bachelor of Science
Computer Engineering

Royal Military Academy

(No Degree)
Robotics

Oxford University

(No Degree)
Marketing

Projects

A Data-Filled Day with the Mouse

Researched how the Disney Blog could prove useful to the corporation by determining popular topics, prime time for blog posts, and negative comment detection (sentiment analysis).

IPython Notebook

Statistical Analysis of the MBTA

Using data from the MBTA’s fare collection system, my team and I focused on developing ridership prediction models. We explored the more nuanced aspects of when Bostonians ride the T, such as sporting events, weather, and geographic location of subway entrances.

Project Website

Pitch Sabermetrics

Similar to the popular movie Money Ball, my team and I carried out an investigation into how Major League Baseball coaches could carry out pre-game analysis. The analysis could be used to gain insight into how an opposing pitcher might change his strategy based on the batter, inning, and other game features.

The Code

Predicting Hospital Staff

Accurately predicting Emergency Room attendance can alleviate hospital-staffing dilemmas and reduce wait time for patients. In this scenario, I applied various Machine Learning techniques to emergency room visit records and historical event data in order to predict admissions one week in advance.

The Code

Improved Graph Bridge Cutting

The BridgeCut algorithm is a very effective technique for producing clusters in networks. However, there are cases in which the original algorithm proves inefficient. In order to solve this issue, I introduced three improvements to the BridgeCut algorithm.

Full Report

Real-time Realistic Path Planning

A primary objective for mobile robots is to autonomously navigate unknown environments. As part of my undergraduate thesis, I presented methods in which smoothed paths and curved turns can be achieved, to make movement more efficient and aesthetically pleasing.

Full Report

Massive Cloud Analytics

Using large clusters, I developed Hadoop MapReduce jobs to analyze terabytes of data for the Department of Defense. The results were pipelined directly to analysts, providing previously unavailable intelligence.

Protein Structure Detection

Protein Folds are responsible for the 3D structure of protein. Using spectral analysis algorithms, I attempted to find similarities between protein folds in order to predict functionality for unknown proteins.