Mathematics and Inference at Your Service

As world-leading independent PyTorch and machine learning experts, you can count on us to get top notch expertise and service.
We enjoy working both with smaller and medium enterprises and - for trainings and targeted work - larger corporations.


Growing your Expertise

You want to start with Artificial Intelligence? You want to get better results faster? You need the best possible no-nonsense training for your experts? We help you get up to speed fast and can help you stay au courant with the latest developments in the research and open source community.

Our training workshops are among the most in-depth and solution oriented offerings for PyTorch and AI topics. We focus on the topics important for your next projects. This way, you get off to the ideal start, and have immediate initial results.

Naturally, we are happy to do the next steps with you, too.


We can support you at all stages of your AI projects or deliver from data to prototype application.

Getting started

With our world-class training, help in tool selection and model design, we help you with the first steps of your AI project. Combining the leading open-source tools, a good overview of current developments, and a deep mathematical and statistical know-how, we help you make most of your data.

Model Analysis and Improvement

We can provide a fresh view on your modelling and highlight enhancements to your models.

We also help you open the lid of the AI blackbox: we help you attribute results to their sources.

Deployment and Optimization

We help you to deploy your models. From server to edge devices, we help you get your model to your users. We can also help beyond the technical side of deployment: Explaining your AI to stakeholders is just as important as putting out the right results.

We can also help make your models run blazingly fast. We have a track record of effective and innovative optimizations both in open source projects like PyTorch and on custom modelling code.

Open Source Support

Professional-Grade Support

While many of the Open Source projects relevant to AI practicioners offer excellent public support, sometimes you need a dedicated point of contact. We can provide first class private advice, answers and reviews affecting involving your non-public code, and guarantee timely responses.

Furthering Open Source Solutions

As one of the leading independent contributors to Pytorch, we leverage our in-depth knowledge to fix the bugs bothering you or propose features you desire.

By its open source nature, we cannot guarantee inclusion to PyTorch. Our track record shows that we are successful in aligning our solution development with the community desires.

Tailored Results and Building Know-How

I is important to not reinvent the wheel – but sometimes, the ready-made solutions are not fit for the purpose. Just as tailoring solutions to your business is key, it is crucial to build up own know how in your core business instead of depending on external services.

Using modern tools, we efficiently jointly develop high quality, maintianable bespoke software. In our collaboration, you have the opportunity to build and grow your own know-how.


Why MathInf exists

  • Our customers deserve excellent, honest consulting that helps them in immediately and in the long run.
  • We stand for taking delight in our work.
  • We aim for purpose and sustainable success - with our clients, as a company, and in the community.


Dr. Thomas Viehmann

Thomas Viehmann is a modeller, mathematician and consultant.

Thomas has been helping large and small firms with their models for more than 10 years. In 2018 he founded MathInf GmbH to better focus on servicing his clients with the best machine learning training and consulting.

Thomas is a PyTorch core developer and well-connected in the applied AI community. He studied mathematics and computer science in Bonn, receiving a Ph.D. (Dr. rer. nat.) in mathematics in 2009. Thomas worked as a certified actuary with the German Actuarial Society. In addition to his PyTorch and machine learning trainings, he also taught mathematical risk management with the German and European actuarial academies 2015-2020.