Iguazio boosts Azure Machine Learning with MLOps automation, enabling companies to rapidly deploy AI

Data Science company Iguazio offers an integrated MLOps platform designed to automate and streamline machine learning pipelines. By accelerating and simplifying the development, deployment, and management of AI applications, the Iguazio platform enables data scientists to focus on delivering better, more accurate, and more powerful solutions—instead of spending their time on infrastructure and data wrangling. 

Getting to market faster

When hygiene solutions provider Ecolab needed a data science collaboration platform that would enhance its customer offering, Iguazio stepped in to help. “Ecolab was looking to develop advanced predictive risk models for its customers, using sensor data,” explains Yaron Haviv, Co-Founder and CTO of Iguazio. “They were looking to streamline the entire process, from data preparation and model development to roll-out of operational pipelines into production.” 

“Prior to 2019, commercial machine learning models followed a ‘rewrite-and-deploy’ pattern, where model development occurred independent of the application developers,” explains Gregory Hayes, Ecolab’s Data Science Director. ”This approach led to roll-out timelines exceeding 12 months on average. We wanted to simplify the process of provisioning scalable and cost-effective compute resources but also needed to reduce the timeline from proof-of-concept to production deployment.” 

Ecolab was certainly not alone in the hurdles it faced. “The process of getting data science from lab to production—and generating real business value with AI—is very complex and requires a lot of resources,” explains Sahar Dolev-Blitental, Vice President of Marketing at Iguazio. “This is why companies like Ecolab can spend a year or more getting one model to production. They have to manually do all of the work of piecing together different technologies and open-source tools. They often don’t have a way to collaborate across teams and introduce automation to simplify and speed-up the process. This is a common problem."

Streamlining data science end-to-end

“Ecolab reached out to us for a trial which enabled team members to work collaboratively, significantly reducing the time to deploy new AI applications, and at a fraction of the effort required previously,” explains Haviv. Progress was fast. The Iguazio solution, built on an open-source architecture that includes Kubeflow, was designed to facilitate machine learning operations. This made it easy to adopt, and helped Ecolab extend and simplify the process of collecting and analyzing data. Robust feature engineering, AutoML and model deployment capabilities are unified together under a single automated CI/CD workflow, enabling rapid development and deployment of production pipelines.

“Soon, data engineers, data scientists, application developers and product owners were all working together using the Iguazio-Microsoft solution,” reveals Haviv. “They were automating their machine learning pipelines, continuously developing, integrating and deploying data and AI products, and scaling their workloads horizontally. All these significantly reduced execution time. It was exciting to be able to help Ecolab realize their ambitious plans for new AI applications.”

“What we enable with our solution for companies like Ecolab is to break the silos between data scientists, data engineers and MLOps engineers,” explains Dolev-Blitental. “It can be very frustrating for a lot of companies who are trying to build this from scratch in-house. We offer a single end-to-end solution, which facilitates the entire production workflow.” 

Ecolab had more than 20 data scientists working on its solution, which is a potential cause of internal roadblocks. “Namely, if you have a large team of data scientists, how do they collaborate?” asks Dolev-Blitental. “How do they standardize and share features? How can they introduce automation to reduce workloads?” Iguazio realized this needed to be addressed.

Instead of having each data scientist separately working on data collection, preparation, feature engineering, creating the feature vectors, Iguazio’s customers can have one central place where they can collaborate, build and share features seamlessly. “Having a single platform with an integrated features store also means feature standardization across the organization, which in turn improves model accuracy, because everyone is using the same methodology and features,” notes Dolev-Blitental.

Accelerating production times

In a single year, Iguazio’s platform empowered Ecolab to accelerate its production times five-fold, breaking down silos between data scientists, data engineers and DevOps practitioners. “It was a huge win,” says Hayes from Ecolab. “Leveraging Azure DevOps and Iguazio, we were able to integrate the creation of machine learning models into the Agile development process. This allowed our data scientists and machine learning engineers to work side-by-side with the product development teams, which improved cross-functional collaboration. Ultimately, these gains enabled our teams to more quickly build and deploy models at scale to meet business requirements.” 

The Iguazio platform is open and deployable anywhere—multi-cloud, on-premises, or edge. “It really makes the whole data science process simple for customers,” enthuses Haviv. “There is tight and broad integration with different Microsoft products, including Azure DevOps, Azure ML, Azure Compute, Storage and Kubernetes (AKS), Azure Data Lake, Azure Synapse Analytics, Azure Event Hubs and other databases and storage. This means that customers who are using the Azure cloud, and who want to accelerate their deployment, can do it in a completely seamless way.”