About Me
I am currently part of the AI Team at idea Engineering (Axel Springer group) where we work on machine learning and deep learning projects covering computer vision, natural language processing, natural language generation, text-to-speech, speech-to-text and many other interesting topics.
My educational background is in theoretical and computational mathematics, Bayesian learning and classical machine learning. Over the last few years I had the privilege to nose-dive into the deep learning domain and do interesting research. Still loving all the input I receive from this incredible field and can not wait to give my contribution.
I am proficient in Python, thanks to which I forgot all the C++ learned when doing numerical approximations of PDEs #noregrets. I have have worked with both Tensorflow and PyTorch, love both but the most love goes to PyTorch. AWS, Docker and Jenkins are also things I had the chance to deal with.
I am passionate about politics, ethics, phylosophy and the dynamics of society in general. Despite not knowing anywhere near enough about any of this, I love having coversations on these topics. Or on any topic really.
When not in front of a laptop, you will hopefully find me throwing frisbees in a park.
Open-source projects
Implementation of a non-autoregressive text-to-speech Transformers network.
Samples, fully trained models and documentation are available on the project page.
Based on a server-less AWS architecture, this model (German, not open-sourced) is used in an experimentalfeature on the welt.de website for synthesis of audio articles.
Experience
In the AI team of ideas Engineering I mostly work on generative models, neural speech synthesis and NLP.
Here I contribute in driving the research and implementing MVPs and some ML backend infrastructure.
At idealo I was part of the early small Machine Learning team which mostly tackled computer vision topics such as Image Estetics, Image Deduplication and Image Super-Resolution. Super Resolution was my main project during this time.
I also dealt with data pipelines and deployment and I had the chance to present our results to the ML community at conferences and meetups.
The technologies I used here include Python (keras, tensorflow, pandas, sklearn, ..), Docker, AWS (EC2, S3) and Openshift.
Education
Technical University Berlin
MSc Mathematics
2014 - 2017
Applied mathematics for scientific computations, engineering and applications.
Main focus on Optimization Algorithms, Statistics and Numerics for PDEs, later specialized in Machine Learning and Bayesian Inference.
In my thesis I defined a novel Bayesian Optimization approach to jointly optimize closely related black-box functions using Gaussian Processes.
During my research assistant experience I also created a parallel version of the (collapsed) Latent Dirichlet Allocation algorithm for natural language topic modelling..
University of Parma
BSc Mathematics
2011 - 2014
Talks
April 2021: ODSC East, Remote, Brand Voice: Deep Learning for Speech Synthesis
June 2019: PyData Amsterdam, Amsterdam, Low to High Resolution: a walk through an Image Super-Resolution project with CNNs and GANs [video]