Track2Seq - LSTM Recommender
Card image cap

Technology: Python, TensorFlow, pandas, numpy
Tags: Machine Learning, Data Engineering, Deep Learning, Recommendation Systems

Automated playlist continuation has been coined as one of the grand challenges in music recommendation. For my Computer Science master's thesis I decided to tackle this problem by leveraging an approach from natural language processing. The idea is to create a language model based on song or track URIs. The resulting recommendation system is called Track2Seq. Initial experiments on a large scale dataset show an over 24% improvement in r-precision when compared against state-of-the-art recommendation baselines.

Teaching A Simulator How To Drive
Card image cap

Technology: Python, OpenCV, Keras, Tensorflow
Tags: CNN, Computer Vision, Autonomous Cars, Deep Learning

One center-piece in the quest for autonomous cars is called behavioral cloning. By turning human driving in machine readable information it is possible to train a computer to learn to steer a car in the correct direction. While a lot of different approaches have been undertaken to solve this task, a very promising avenue that has been developed in the last decade is deep learning. Deep neural networks seem to have great capabilities to act on image data. In this project a convolutional neural network (CNN) is created that is capable of driving a vehicle in a computer simulation without human intervention. The final model scores a Mean Squared Error of .0176.

Digits Classifier
Card image cap

Technology: Python, TensorFlow, Django
Tags: CNN, Machine Learning, Computer Vision, Deep Learning

In this project we're training a convolutional neural network to recognize handwritten digits from scratch. Instead of using data that has been made available by other researchers, we'll be using our own hand drawings. With as little as 1010 hand-drawn images the model is able to classify new unseen images with an accuracy of over 96%. The final model can be tried out here.

Social Impact
Card image cap

Technology: Python, Django, d3.js
Tags: Statistics, Analytics, Dashboard

Social Impact is an analytics dashboard to measure social engagement. Developed for a content outlet it enables users to measure the impact of content release on social media. Engagement metrics on a variety of social media channels are collected and presented in aggregations for several campaigns. Doing so automates the reporting mechanism and offers an overview of the total impact over all campaigns.

Imputation Module for Python
Card image cap

Technology: Python, sklearn, pandas
Tags: Machine Learning, Statistics, Data Engineering, Analytics

Impyte is a lightweight Python module that has been created with one goal in mind: Simplify the way a researcher deals with missing values. Impyte leverages machine learning algorithms to complete missing information and offers intuitive visualization methods to help with data engineering and analytics.

Predicting Why Employees Leave
Card image cap

Technology: Python, sklearn, pandas
Tags: Machine Learning, HR, Clustering, Random Forest, PCA

Measuring employee satisfaction is a tough and highly complex task. There are a lot of different dimensions in play and turning them into quantifiable format, less to say machine-readable information, can pose a challenge. In this project a model is being used that creates employee clusters to enrich the Kaggle HR dataset and in turn predict which employee is likely to leave the company.