.png)
Reproducible
Machine Learning in production
Using Machine Learning Operations (MLOps) we provide a process to design, build, train and manage reproducible and testable Machine Learning software.
The following steps define our process of working
Developing machine learning application starts with identifying the scope of the ML-application, the success criteria, and a data quality verification. The goal of the first phase is to ensure the feasibility of the project and define realistic goals.
The second phase of the process aims to prepare data for the modelling and training phase. Data selection, data cleaning, feature engineering and data standardization. Our preprocessing code follows the standard software design principles. We write clean code that is unit-tested to ensure reproducibility and avoid bugs
This phase aims to specify one or several machine learning models to be deployed in production. This includes model selection, model specialization, model training and model evaluation. Since ML-development is an iterative process we ensure an infrastructure that can be used by one data scientist, or a team of data scientists to train complex models and track model performance over different data and model versions.
After training and evaluating a successful model, the model can be deployed on premise, in the cloud or hybrid. Depending on the architecture and data-load. The proper hardware will be selected. Model deployment includes the following tasks: inference hardware selection (CPU, GPU or TPU), model evaluation in production, providing a back-up plan for modal outages and a deployment strategy.
To prevent model performance drop caused by Data Drift or Concept Drift the model is continuously evaluated to decide whether the model needs to be retrained.
Join the team that spearheads the MLOps revolution
and deploy State of the Art deep learning solutions
ReBatch is committed to protecting and respecting your privacy, and we’ll only use your personal information to administer your account and to provide the products and services you requested from us. From time to time, we would like to contact you about our products and services, as well as other content that may be of interest to you