Sebastian Wagner

AI Engineer @appliedAI


    •

Hi, I'm Sebastian, an AI Engineer at the appliedAI Initiative in Munich. My interests are Machine Learning, Natural Language Processing and Computer Vision.

During my MS Computer Science at Ludwig-Maximilians University in Munich, I worked together with Hinrich Schütze to learn better cross-lingual word embeddings for Chinese and English.

Previously, I've worked with Eirini Ntoutsi on mining opionated data streams and published a paper on it at the European Conference on Machine Learning & Principles and Practice of Knowledge Discovery in Databases.

Besides my interest in improving learning algorithms, I also find music a fascinating subject to study. In Spring 2015, I interned at Carnegie Mellon University's Computer Music Lab and worked with Roger Dannenberg on automatic music accompaniment.

Together with Georgia Olympia Brikis, I co-founded the weekly AI Journal Club at LMU.

From my blog

Reinforcement Learning and Supervised Learning: A brief comparison

Should we be afraid of artificial intelligence?

Selected projects

Flink Cluster Model Trainer

This project was conducted as part of the Big Data Innovation Lab at our university. The idea of this project was to parallelize hyperparameter tuning for ML algorithms. Users select Machine Learning models (Linear Regression, SVM, Neural Networks) and train them on a Flink cluster.

Transfer Learning for Facial Expression Classification

Class project for 'Deep Learning for Computer Vision'. Trained different ConvNet architectures to perform well on FER 2013 dataset from Kaggle. Beat open-source baselines and got accuracy close to SOTA.

Traffic flow prediction for KDD Cup 2017

Team submission to KDD Cup 2017. Predict traffic flow from time-series data. Involved mostly data preprocessing, feature selection and modeling (SVM, LR, Neural Networks). Scored top 10% of participants.

iOS Chatbot for Ridesharing

Weekend project that uses an IBM Watson chatbot to book shared rides between cities. Users are matched based on geolocation. Involved mobile and backend development.