TrueLayer Deploys Dotscience MLOps Platform to Speed AI Development and Reduce Engineering Effort and Risk from Untracked Models

Europe’s Leading Provider of Financial APIs Deploys Dotscience on AWS to Enable Reproducibility, Provenance and Metric Tracking of AI Models

Dotscience, the market leader in DevOps for Machine Learning (MLOps), announced leading financial API provider TrueLayer has selected the Dotscience MLOps platform to improve reproducibility, model and data versioning and provenance tracking in response to demand for improved productivity, collaboration, governance and compliance of AI initiatives. The solution will be used to help TrueLayer deliver models faster and more safely into production through controlled collaboration and the integration of Dotscience into both TrueLayer’s prototyping and production pipelines for ML. With Dotscience, TrueLayer will also be able to track back from a model running in production to exactly which dataset and parameters the model was trained with and where the data originated from.

“Implementing a platform that includes data versioning and model provenance with other tools requires piecing together products from multiple vendors,” said Luke Marsden CEO and founder at Dotscience. “Dotscience helps innovative companies like TrueLayer achieve reproducibility, accountability, collaboration and continuous delivery across the AI model lifecycle by providing an end-to-end platform with these requirements baked in, while also integrating as needed with the company’s version control and CI systems. This enables data science and ML teams to own and control the entire model development and operations process while fitting into the company’s existing DevOps framework.”

Read More: Augmented Reality is the Future of Healthcare Industry

Dotscience’s deployment for TrueLayer was completed in October 2019. Since then, TrueLayer has deployed multiple ML models into production and uses Dotscience to keep track of the models both in the rapid, iterative prototyping phase using Jupyter notebooks, in production data and model pipelines, and when using a grid search to explore the space of possible hyperparameters to optimize ML models.

Dotscience Platform Serves as Foundation for TrueLayer MLOps Best Practices

TrueLayer has deployed Dotscience in a private deployment on Amazon Web Services (AWS), and via an integration with S3, GitHub and CircleCI, uses Dotscience for prototyping and experimentation of models in Jupyter and deploying the tested models into production. TrueLayer’s ML team has created a best practice MLOps platform with Dotscience at its core, using a unique approach which aligns with different phases of the model development lifecycle.

Read More: World’s Largest Blockchain Application Competition ‘Klaytn Horizon’ Winners Announced

“The ability to derive business value from AI is key for TrueLayer. Using Dotscience, we deployed our first ML model into production in just a week and onboarded new data scientists quickly, accelerating time to market, reduced overall engineering costs and reduced risk from untracked models,” said Luca Palmieri, ML and Data Engineering at TrueLayer. “When we are developing a new ML model, or exploring new approaches for an existing model, rapid prototyping is done within a hosted Jupyter environment available directly within Dotscience. This saves our data scientists time having to set up their own ML development environment and enables better tracking of progress during this rapid prototyping phase, liberating our team from the costs and risks associated with manual tracking.”

Once the model approach is narrowed down in the Jupyter environment, code from the Jupyter notebook is factored out into a version controlled Python library in GitHub, which can be accessed via Dotscience’s built-in GitHub integration. Read more about how TrueLayer created an MLOps pipeline for Fintech with Dotscience with the case study on the Dotscience blog.

In separate press releases today, Dotscience announced it has gained momentum in the MLOps ecosystem with new integrations, technology partnerships, product innovations and expanded multi-cloud support and a native GitLab integration making MLOps available to a growing GitLab community.

Read More: Artificial Intelligence In Restaurant Business