Felix Effenberger, Developer in Stuttgart, Baden-Württemberg, Germany
Felix is available for hire
Hire Felix

Felix Effenberger

Verified Expert  in Engineering

Mathematics Developer

Stuttgart, Baden-Württemberg, Germany
Toptal Member Since
May 31, 2019

Felix is a mathematician (Ph.D.) and computer scientist by training, turned researcher in neuroscience and machine learning, and then turned entrepreneur (co-founder and former CTO at SF-based machine learning stealth startup). He currently works from Europe as an advisor to his own startup and as a freelance researcher, software engineer, data scientist, and trainer.



Preferred Environment

Git, Python, Linux

The most amazing...

...thing I've worked on was the development of a high speed real-time lossless + lossy video compression framework on GPUs compressing several GB/s.

Work Experience


2019 - PRESENT
Stealth Silicon Valley Startup, San Fracisco, CA, USA
  • Developed context-adaptive video coding framework for H.264 and H.265 in C, C++ driven by machine learning algorithms (PyTorch, TensorFlow).
  • DevOps and architecture management for development and experimentation platform (Azure, Terraform).
Technologies: Terraform, Azure, DevOps, Git, TensorFlow, PyTorch, Deep Learning, Scikit-learn, Machine Learning, NoSQL, C++, C, Python

Freelance Data Scientist

2019 - PRESENT
nextbike GmbH, Leipzig, Germany
  • Contributed to a project for a shared mobility provider, specifically a bike sharing system. Built Python-based forecasting framework predicting the expected number of rentals and returns for each zone of a city in real time.
  • Developed statistical methods for forecasting rental and return numbers per zone.
  • Built predictive rebalancing framework, i.e. system deciding which bikes have to be moved from where to where for optimal service availability of the bike sharing system that calculates optimal tours for the service personnel (traveling salesman problem with multiple vehicles, as well as pickup and delivery constraints).
  • Built REST API for the above frameworks in Python, using Python Flask framework.
  • Integrated this solution into the client's infrastructure.
Technologies: REST APIs, Scikit-learn, Flask, Python


2017 - 2019
Stealth Silicon Valley Startup, San Fracisco, CA, USA
  • Co-founded and raised angel money and seed round (at $13.5 million valuation).
  • Researched neuroscience-inspired signal processing techniques with a focus on image and video compression. Responsible for everything tech, managed team of five engineers. As scrum master, chose technologies and set coding standards, managed cloud infrastructure, did code reviews. Oversaw deep dives into engineering problems where necessary.
  • Developed cross-platform (macOS, Linux) error-resilient video conferencing system tolerant to high levels of packet loss and changing channel conditions in C++.
  • Developed cross-platform (macOS, Linux) high speed lossless and lossy ultra-low latency video compression algorithm (>2GB/s) on GPUs via OpenCL, with C++ wrappers.
  • Developed machine learning techniques for the semantic analysis of video signals (using scikit-learn, pyTorch, TensorFlow).
  • Developed custom H264 and H265 codecs.
  • Developed a framework for managing cloud infrastructure on AWS and Azure via Terraform (infrastructure as code).
Technologies: Amazon Web Services (AWS), Terraform, Microsoft Azure, GitLab, OpenCL, Objective-C, Assembly, C++, C, Python

Postdoctoral Researcher

2015 - 2017
Frankfurt Institute for Advanced Studies, Frankfurt, Germany
  • Served in a research position in neuronal morphology modeling and data analysis.
  • Published several research papers in the field of neuronal morphology, see https://scholar.google.de/citations?user=IpKlv7kAAAAJ.
  • Co-developed open source Matlab toolbox "TREES toolbox 2", https://www.treestoolbox.org.
  • Supervised MSc and Ph.D. students.
  • Attended international conferences on computational neuroscience, invited talks at Harvard Center for Brain Research, Redwood Center for Computational Neuroscience, Berkeley, and others.
Technologies: Git, LaTeX, MATLAB, Python

Backend developer, part time

2013 - 2015
modelogiq GmbH, Frankfurt, Germany
  • Co-founded the fintech startup; worked as a Python and Clojure back end developer, and JavaScript front end developer.
  • Developed tools for automated financial modeling and spreadsheet generation.
  • Developed a beta version of data-driven JavaScript based front end.
Technologies: Go, Clojure, Python

Postdoctoral Researcher

2011 - 2015
Max Planck Institute for Mathematics in the Sciences, Leipzig, Germany
  • Researched in mathematical neurobiology and computational neuroscience, focused on processes of self-organization in cortical neural networks and the fundamentals of learning (synaptic plasticity). Modeling and analysis of spiking neuron data.
  • Taught courses in computational neuroscience and high-dimensional data analysis.
  • Developed open source Python package hdnet, for high dimensional data analysis using Ising models, https://github.com/team-hdnet/hdnet.
  • Published book chapter on information theoretic methods for data analysis, https://link.springer.com/chapter/10.1007/978-1-4614-8785-2_5.
  • Published several research papers in the field of computational neuroscience, see https://scholar.google.de/citations?user=IpKlv7kAAAAJ&hl=en.
Technologies: Git, LaTeX, Python

Freelance Software Engineer

2013 - 2014
nextbike GmbH, Leipzig, Germany
  • Developed a data-driven Android application for service staff of bike sharing service.
  • Developed a data-driven Android application that allows service personnel to see system status and to perform service and rebalancing operations.
  • Advised on IT infrastructure (replication, fault tolerance, etc).
Technologies: Android, Java

Ph.D. Level Researcher

2007 - 2011
University of Stuttgart, Germany
  • Researched in fields of discrete topology, geometry and combinatorics. Grant by the German Research Foundation (DFG), Project Ku 1203/5.
  • Authored open source software simpcomp, https://github.com/simpcomp-team/simpcomp.
Technologies: Subversion (SVN), GAP, Perl, LaTeX

simpcomp - a GAP Toolbox for Simplicial Complexes

simpcomp is an open source extension (called "package") for the computational algebra system GAP (https://www.gap-system.org/index.html), a well-known system among mathematicians in the field of computational algebra research.

simpcomp underwent formal peer review and is included in every standard installation of the GAP system (see https://www.gap-system.org/Packages/packages.html).

The package enables the user to compute numerous properties of (abstract) simplicial complexes, provides functions to construct new complexes from existing ones and an extensive library of triangulations of manifolds.

simpcomp is free software. The code is released under the GPL version 2 or later (at your preference).

Joint work with Jonathan Spreer.

TREES Toolbox 2 - an Open Source Matlab Toolbox

The TREES toolbox is an open source Matlab package that provides tools to automatically and manually reconstruct neuronal branching from microscopy image stacks and to generate synthetic axonal and dendritic trees. It is a tool to edit, visualize and analyze dendritic and axonal trees and includes methods for quantitatively comparing branching structures between neurons.

Version 2 of the toolbox introduces an object-oriented paradigm, as well as APIs to directly obtain reconstructions of dendritic trees from popular databases such as NeuroMorpho.org.

Joint work with Hermann Cuntz, see https://www.treestoolbox.org/authors.html.

hdnet - Hopfield Denoising Network

hdnet stands for Hopfield denoising network. It is an open source Python package for analysis of parallel neural population spiking data, i.e. parallel spike trains, and can also be used for unsupervised clustering of arbitrary high-dimensional, noisy, binary data.

In particular, it provides a novel method for finding and extracting salient low-dimensional representations of the dynamics of populations of spiking neurons based on a denoising approach to spatiotemporal patterns (STP) contained in the data. Using Hopfield networks trained with minimum probability flow (MPF), the occurring raw spatiotemporal patterns are grouped into clusters of similar patterns in an unsupervised way, assigning to each cluster a memory (the fixed point of the Hopfield dynamics in each cluster).

The package ships with a tutorial and is documented.

Joint work with Christopher Hillar.


Python 3, Python 2, C, C++, JavaScript, Bash Script, Python, Assembly, Assembler, SQL, PHP, Java, GAP, Perl, Clojure, Objective-C, Swift, R, Go, Rust, Kotlin


OpenCL, Django, Flask, Chainer, Hadoop, Angular, Spark, Caffe


Scikit-learn, Pandas, NumPy, Matplotlib, PyTorch, TensorFlow, SQLAlchemy, Keras, OpenGL, OpenGL ES, REST APIs, jQuery, React, Vue, Node.js, TensorFlow Deep Learning Library (TFLearn), Theano


LaTeX, MATLAB, Git, Terraform, Azure Machine Learning, TensorBoard, Seaborn, GitLab, Ansible, Subversion (SVN), Google Kubernetes Engine (GKE)


Object-oriented Programming (OOP), Agile, Functional Programming, Scrum, DevOps, Continuous Integration (CI), Microservices


Linux, MacOS, Amazon Web Services (AWS), Android, NVIDIA CUDA, Windows, iOS, Amazon EC2, Azure, Docker, Kubernetes, Apache Kafka, Blockchain


Machine Learning, OpenCL/GPU, Deep Learning, Mathematics, Mathematical Analysis, Mathematical Modeling, Clustering Algorithms, Artificial Intelligence (AI), Artificial Neural Networks (ANN), Blockchain Development, Embedded Software, Infrastructure as Code (IaC), Google Cloud Machine Learning, Statistical Analysis, Statistical Modeling, Computer Security, Open Source, Torch, Microsoft Azure


MongoDB, MySQL, PostgreSQL, NoSQL, Google Cloud

2007 - 2011

Ph.D. (Summa Cum Laude) in Mathematics

University of Stuttgart - Stuttgart, Germany

2002 - 2007

Master of Science Degree with Distinction in Mathematics, Computing Science

University of Stuttgart - Stuttgart, Germany

Collaboration That Works

How to Work with Toptal

Toptal matches you directly with global industry experts from our network in hours—not weeks or months.


Share your needs

Discuss your requirements and refine your scope in a call with a Toptal domain expert.

Choose your talent

Get a short list of expertly matched talent within 24 hours to review, interview, and choose from.

Start your risk-free talent trial

Work with your chosen talent on a trial basis for up to two weeks. Pay only if you decide to hire them.

Top talent is in high demand.

Start hiring