Introducing Mantis


Mantis AI


github share icon

In this blog post we introduce Mantis NLP, a remote first company we have founded to help companies put impactful data science into production. Mantis has a focus on Natural Language Processing (NLP), but we also want to help organisations more generally with getting their Machine Learning models out of the experimental phase, and into production.

We’ve got about 15 years of experience in data science behind us, and have worked in a number of sectors (academic, legal, health, environment) and for a range of different organisations (government, startups, charities, foundations). These experiences have shaped the way we approach our work in terms of our ethics and work philosophy. We believe that:

  • Experiments should be easily reproducible. This encourages more trust in the results, enables better collaboration within a team, and reduces duplication of work.
  • Data reflect human biases. AI applications mimic those biases and can cause unintended harm. Detecting and reducing ethical risks is the responsibility of all data scientists.
  • AI is not a neutral technology. Applying AI tools for military, political or marketing purposes has an impact on our democracy and free will. We think carefully about the impact our work might have, and are selective about the industries and clients we work with.
  • Open source software and practices have made, among other things, the AI revolution possible. They accelerate progress, and encourage collaboration between people and organisations. We use open source by default, and contribute to the open source community where possible.
  • It’s our responsibility to combat climate change to support a more sustainable way of life. Training large deep learning models can have enormous carbon footprints so we avoid using such models where possible, and use computing resources responsibly and efficiently.

This is not an exhaustive list, we will keep adding as we go. We also aim to explain more about our thinking behind those statements in future blog posts.

We’re inspired by great companies like and Hugging Face, and like them we aim to be active members of the open source community.

We’ve just launched our homepage at, but the best place to keep up to date with what we are up to is through our medium and twitter channels.

MLOps for Conversational AI with Rasa, DVC, and CML (Part I)


Are you interested in working with us?