Technical Skills
My Harvard coursework focused on Python but also included C/Unix and Java. Data Science work included extensive work with the popular python packages for Machine/Deep Learning and data management.
The projects linked below highlight many of the skills I learned at Harvard.
Languages, Frameworks, Libraries and Methodologies
  • Python
  • Open CV
  • Pandas
  • Numpy
  • SciKit-Learn / Statsmodels
  • PyGam / PyMC3
  • Keras / Tensorflow
  • Bokeh
  • Django
  • Flask
  • Docker
  • Agile
  • UML
  • Flask
  • C / Unix
  • Java (incl. Maven)
  • Springboot
  • Microservices
  • RaspberryPi
ML, Deep Learning and Statistical Modeling
  • Linear and Logistical Regression
  • Generalized Additive Models (GAMs)
  • Basic Reinforcement Learning
  • Bayesian Modeling
  • Splines
  • Clustering
  • Non-parametric Models
  • Ensemble Methods
  • NLP (Embeddings, RNN, LSTM)
  • CNN / Mask RCNN
  • Autoencoders
  • PCA
  • Generative Adversarial Networks (GANs)
  • Transfer Learning
Traffic Monitor

Track Detected Objects in Video Stream

By: Mark McDonald
The Traffic Monitor application will detect objects in a video feed and log, report and chart instances of objects that are recognized. A variety of objects can be selected for tracking as well as the source of the video stream.
Traffic
Fashion Recommendation

Identification of Fashion Objects in an Image

By: Mark McDonald, An Hoang, Vivek Bhatia
Fashion-related items are segmented from images using Mask RCNN. Segmented objects are identified based on a large dataset provided in a Kaggle competition. An existing implementation of the Mask RCNN model is modified to apply to this project. This project was the result of a group effort for my course in Computer Vision.
Fashion
EU Green Deal Reporting

EU Climate Reporting for EU Commissioners

By: Mark McDonald, Alan Martinson, Carly Gloge, Hemant Bajpai, Pritam Dey, Taylor Meyer
Harvard Software Engineering degree capstone group project. Our team was tasked by an EU Commission department to build a template which consolidated complex and diverse sources of data into an easy-to-understand front-end designed for EU Commissioner staff members. The effort focused on air pollution.
EU
Mark's Covid Tracker

Daily US COVID-19 Trends

By: Mark McDonald
Various interactive charts are presented to the user show pertinent country-wide, state and county-level trends in the spread of COVID-19 in the US. This project uses Bokeh to build visualizations, Django to host the web pages, and JavaScript for user interactivity.
Mark's
NYC Capital Projects

Predicting outcomes for NYC capital projects

By: Mark McDonald, Michael Sedelmeyer, An Hoang
Final group project for Advanced Data Science class at Harvard. Data available at the beginning of various large NYC capital projects is used to determine how likely the initial project budget and timeline will be after 3 years. This project compares various prediction methodologies as well as data engineering to find a latent reference class for the projects.
NYC
Chatterbox

Barebones web-chat with no fuss

By: Mark McDonald
A simple chat application demonstrating basic HTML, CSS, Flask and Internet messaging using sockets.
Chatterbox
Pathfinder

Off-Road Path Navigation

By: Mark McDonald
Using Keras and OpenCV, Pathfinder will highlight a navigable path on a video of off-road terrain. The model was trained on a small set of publicly available data that was augmented to produce a useful set of training data.
Pathfinder