Back1 Back2 Back3 Legal-Pythia

AI language analysis app for an innovative startup


AI language analysis app for an innovative startup

“Having a partner who is really good at it, who knows his job and can also provide a different perspective, helped us a lot in the end.”


Jeremy Bormann, Founder, Legal-Pythia


-> Read full interview

Project key data

2 people

4 months

AI development environment, AI application for analyzing legal texts

AI language models, LLM/Transformer


mantik (our proprietary AI development platform), S3-Buckets, FastAPI

Why we like showing this case

AI offers young companies a fantastic field of opportunities with new business models. Investments in AI-based business logic are often not as high as many think. The right foundation of expertise and a state-of-the-art IT infrastructure ensure fast development cycles and a short time-to-value. For example, we were able to provide Legal-Pythia with a deployment of the jointly developed large-language model in just one week.

We were also particularly happy to support Legal-Pythia because their explainable AI approach ensures transparency and proves that AI does not have to be a fearsome black box.

Task and solution

The founders of Legal-Pythia had proposed a concept in an ideas competition organized by the City of Munich on how to reliably analyze and compare legal texts with the help of AI in order to

  • improve the administrative service for citizens
  • use taxpayers' money wisely
  • make skilled workers available for higher-value tasks.

The founders had excellent expertise in business law and already had experience in the field of AI. Now they were looking for a partner who could implement the project quickly and according to specifications on a safe technical basis.

We were able to leverage our MLOps platform mantik, which we had developed over the last few years, and our previous experience in AI-based text analysis for this project. This meant that we were able to deliver an interface to the jointly developed language model via automated processes within only a week and with a manageable amount of effort, which won over the City of Munich in the final pitch.

In four months of intensive joint work with the founders, doing several rounds of evaluation, we were able to develop the solid foundation of the Legal-Pythia code into a productive model that can be used for successful document processing via the API .

AI - only ever as good as its infrastructure

There are now many tools that facilitate the development of AI applications. However, the actual performance of an AI development environment - time-to-value, flexibility in the development and training of models, cost efficiency, data security, reproducibility, rapid deployment, efficient collaboration, monitoring and maintainability - often only becomes apparent during the course of the project. If fast scaling, special data security, processing of large amounts of data or heterogeneous data, automatic deployment or end-to-end integration are required, or if the AI application is the main asset for competitive advantage, it is worth making no compromises when it comes to the development environment.

Ambrosys combines AI expertise with the highest standards in software engineering and develops high-performance AI infrastructures according to best DevOps practices.


→ Questions or thoughts on high-performance infrastructure for AI development? Dr. Markus Abel is looking forward to hearing from you.

Go back