The real challenge isn't building an ML model, the challenge is building an integrated ML system and to continuously operate it in production.
Every day new approaches and new tools are released regarding ML in production. We do constant research on these and keep our way of working up to date with the best tooling on the market.
We build ML powered applications. Our chapter consists of multidisciplinary people to cover all phases of an ML project.
We share knowledge with the rest of the delivery organization to make sure our implementation projects are according to best practices for putting ML in production.
Internal tooling helps us to speed up delivery, standardize approaches and technologies and enable efficient use of time and resources. With our internal tool GCCP we are able to reach high internal quality of our solutions without the downside of slow start. GCCP is our command line “Cookie Cutter” tool that can easily replicate boilerplate code for our most used GCP components. GCCP enables boilerplate customization which helps us to adapt the templates for the needs of a specific use case, and then build more custom logic on top. In addition, it generates configuration files for secure infrastructure setup and CICD.
Open source contributions
Taking ownership and responsibility is in our DNA. We are always happy to help the community to improve the tools we use ourselves, thus we are very enthusiastic about contributing to open source software. Have you heard about Connexion? We are convinced of the API-first approach for microservices. It allows us to clearly separate definition and implementation, efficiently collaborate with the stakeholders and reduce applications developing costs. Connexion is a great Python framework and the best choice for our needs. We are very proud to be part of the community and to help move Connexion forward!
Best practices and code quality
Standardization has a positive impact on any business. It enables efficient use of resources, reduces risks and delivery costs. The Software Architecture chapter provides the guidelines, documents and propagates best practices at ML6. CICD and automation in general is part of our development culture. We aim to avoid any manual action when possible which enables delivery of high quality software faster and allows us to focus on solving business problems. We are fans of a code review process. Code review leads to better implementations and more efficient solutions, it helps to find bugs earlier when those are cheaper to fix, as well as to share the experience of senior developers and to improve team cohesion.