7 Real-World Use Cases Of Successful MLOps Adoption

Machine learning plays a crucial role, from generating the best product recommendations as per your requirement to helping you discover the best entertainment show on a streaming channel. Even though every industry strives to benefit from machine learning capabilities, the hard truth is that ML projects fail.

In a 2020 study, Gartner reported that only 53% of AI projects successfully moved from prototype to production — and that’s at enterprises with some AI expertise.

If we research more on ML project failure reasons, these might include – wrong AI strategy, models failing to generalize, deployment hurdles, no appropriate monitoring, and more. But, setting up an MLOps team and processes would help drive continuous improvements and keep every ML project up for long-term returns. So in this article, we explore different successful real-world use-cases of MLOps.

‍

What is MLOps?

MLOps, as its name suggests, combines two terms - machine learning and operations. It defines a set of practices that help standardize, simplify and modernize processes involved in ML applications deployment. It spans the entire lifecycle of ML models, from POC research to model development to post-deployment phases.

MLOps helps unify ML project tasks, streamline the continuous delivery of performant models, and enable effective collaboration between ML and other teams. It helps businesses with accelerated model deployments and better compliance with regulatory standards. MLOps supports cross-functional teams, including C-Suite leaders, DevOps team, MLOps engineers, operational team, and risk-compliance experts. It provides the following benefits:

• A sophisticated way to scale machine learning projects
• A better way to manage, build, deploy, monitor, and scale ML models in a production environment
• Effective risk assessment and compliance management for ML projects
• Ensure better collaboration between teams
• Reproducible ML workflows and models
• Helps plan competent AI strategy that meets business goals  

‍

7 Real-World Use Cases of MLOps

1. Merck research labs accelerates vaccine research and discovery

Merck Research Labs implemented MLOps to accelerate vaccine research and discovery. Dr. Kam Chana, Director of the Scientific Data and Computing Platforms group at Merck Research Labs, speaks about the company’s efforts to drive ML-based innovation in the pharmaceutical and healthcare sector.

The healthcare company faced the following machine learning operational challenges:

• Prolonged research
• Disconnected teams –DevOps and ML team
• Stakeholder buy-in
• Technology mismatch
• Ineffective ML lifecycle
• Technical debt

As a result of these hurdles, the company found that the time required to deploy models has increased, making the ML-driven drug research process expensive. They adopted Algorithmia’s MLOps Platform, which provided them with

• Excellent tech stack to screen millions of virtual compounds cost-effectively and swiftly
• MLOps-driven model publishing to accelerate other workflows with low latency
• Streamlined workflows that allowed teams to focus on research efforts

Merck Research Lab accelerated vaccine discovery and confirmed these results with MLOps implementation:

• Enhanced processing capability to screen compound images automatically
• Streamlined ML operations with precise automation
• Freedom to use IDE of researcher’s choice avoiding technology lock-in

‍

2. GTS data processing company enhances its ecosystem with MLOps

GTS Data Processing is a German IT Company that offers Infrastructure-as-a-Service and Software-as-a-Service platforms to companies across Europe. GTS adopted MLOps to enable its clients to provide compute resources, ensure reproducibility, and significantly accelerate model deployment.

Challenges faced by GTS:

• Need for governance and security
• Challenges in cloud adoption
• Tough data protection standards in Germany that hindered AI project advancements

‍

Solution:

GTS leveraged enterprise MLOps capabilities that helped data science teams to access the compute resources, data, and tools. MLOps implementation improved team collaboration and streamlined governance among processes. They chose Domino Enterprise MLOps Platform to fulfill the following criteria:

• Enterprise-grade security
• Customization
• Self-service infrastructure
• Powerful collaboration and governance

‍

Results:

GTS developed its DSReady Cloud to help companies scale their data science effectively while offering a secure environment and compliance with GDPR. DSReady cloud solution helped bring together tools, technologies, and collaboration capabilities required to manage and scale data science projects. It helped data science teams with

• Faster development of models
• Reproducibility
• Streamlined governance
• Faster deployments

‍

3. EY accelerates model deployments with MLOps

EY is a global financial firm that offers professional services such as tax, assurance, and advising. It harnesses AI and machine learning technologies for anti-money laundering, fraud detection, checking economic compliance, and trade surveillance.

EY wanted to leverage AI against financial crimes with a broader objective of building trust and integrity in the financial market.

EY worked with several fintech companies that took much time to deploy a single machine learning model. They were unable to keep up with the growing financial deceit due to which their models became irrelevant even before productionalization.

EY wanted faster GTM for ML-based finance solutions to reduce financial crime cases. The crucial aspect was cutting consumers’ crime detection timeline from years to hours. For this, they required:

• Openness to any tooling option and frameworks
• Modularity to encourage reusability, code-sharing, and work on components
• On-demand scaling on suitable infrastructure

‍

Solution

EY implemented MLOps to accelerate model deployments using diverse frameworks and libraries. The massive increase in operationalizing models helped EY empower its customers to reduce the rate of financial crimes committed over time.

‍

Key outcomes:

• Accelerated model deployment helped EY’s clients to reduce the rate of financial crimes committed.
• Reduced false positives by 40-60%
• ML pipeline modularity helped drive reusability, versioning, and code-sharing

‍

4. MLOps helping KONUX excel in predictive maintenance operations

KONUX utilizes IIoT and ML to transform railway operations. Their platform helps monitor rail infrastructure and inform any maintenance issues before they grow. It serves some of the largest railway companies, including the railways in Germany, the UK, and China.

KONUX provides a predictive maintenance system with intelligent analytics that can identify trains and railcars on the track from the sensor data and alert on abnormalities. Such predictive maintenance helps ensure enhanced safety, availability, and better lifespan of the existing infrastructure.

The company adopted MLOps to enhance its machine learning experiments. They employed Valohai - an MLOps platform that allowed continuous training of the production models and offered insights into the SaaS product. It helped the KONUX team to run simultaneous experiments within a short timeframe. Automated experiment tracking and orchestration boosted productivity exceptionally.

MLOps implementation allowed the KONUX team to focus on data science. Systematic version control on ML experiments enabled teams to review each component on a need basis. MLOps solution helped them in large-scale experimentation without any costly resources and hurdles.

‍

5. MLOps helped PadSquad to enhance ad performance in real-time  

A mobile software company PadSquad used the Iguazio Data Science Platform to enhance the performance of ads they offer to their global customers. PadSquad implements innovations in media technology to elevate audiences’ engagement and experience.

PadSquad implemented the MLOps platform to improve and optimize ad performance and reduce media costs incurred by their consumers. They were facing the following challenges:

• Difficulties in streamlining the AI development process
• More time spent in operations than the core work
• Expensive in terms of slower GTM

PadSquad implemented a data science and MLOps platform to analyze events and sentiments from real-time online ads. This helped deliver innovative and relevant creatives to the audience. The automated ML pipeline supported the entire ML lifecycle, ensured advanced feature engineering capabilities, and abstracted away DevOps. MLOps implementation helped PadSquad streamline the process of AI applications development from research to deployment.  

The platform helped automate the operational side of developing and deploying ML models and allowed PadSquad data scientists to focus on business logic rather than operations. MLOps platform enabled open and integrated architecture in the entire process, from data prep to model deployments and management.

The main benefits of MLOps adoption by the media company include

• Automation and orchestration of processes
• Faster GTM
• Enhanced ad performance with better experience and engagement
• Focus on data science instead of operations

‍

6. Senko Logistics Group enhanced shipment volume accuracy

Senko Group Holdings, an integrated logistics service partner, provides logistics business as one of the core offerings for the apparel and e-commerce sector in the Tokyo metropolitan area. Senko Group wanted to improve the efficiency of human resource planning with AI-powered shipment volume forecasts.  

The logistics company faced the following challenges regularly:

• Complexities in predicting shipment volumes 
• Requirement of additional arrangements to fulfill shipments
• Challenges associated with customer support and management
• Compliance with Government’s work style reform policy 
• Challenging to use existing tools in operations 

‍

Solution

Senko group chose MLOps adoption to start AI-driven shipment volume forecasts. They integrated H2O data science platform to ensure automated and seamless feature engineering and feature tuning. This MLOps implementation helped the logistics team to streamline the operational procedures. They used SPSS for data processing and a data science platform for predictive modeling and predictions, helping them organize the operational procedure.

‍

Senko group grabbed these benefits with AI adoption: 

• Enhanced prediction accuracy
• Streamlined operations
• Reduction in the workload of staff

‍

7. Starbucks India applied data-driven strategies 

Starbucks is one of the prominent beverage brands with more than 24,000 outlets that delight people with some of the finest coffees in the world. Starbucks India is a leading retail chain in India that created a remarkable market space with its expertise and aggressive expansion strategy. 

The brand was looking for a data solution that would help them double the revenue with minor revisions in offers and discounts. Their focus was on reducing churn with appropriate measures and discovering opportunities for upselling and cross-selling. 

‍

Solution

Starbucks India used an integrated data and analytics platform to collect and analyze loyalty data across different channels. Marketing analysis of micro-segments helped Starbucks India to identify 

• The best customers based on their spending patterns
• Customers that can repeat purchases to ensure revenue

With a data-driven loyalty strategy program, Starbucks India executed marketing tactics to retain customers proactively and devised custom strategies relevant to each microsegment. 

‍

Results:

• Higher revenue in targeted campaigns
• Reduced customer churn
• Inactivity trigger helped in proactive retention of customers

‍

Final Thoughts

So, what is your impression about MLOps now? MLOps is a must-have for modern AI-centric organizations. No matter how ML capabilities are beneficial to modern businesses, reaping the benefits of these ML capabilities is possible only with appropriate talent, practices, and MLOps tools. Hence we tried to introduce a few commending MLOps examples in this blog. Thanks for reading it. I hope this article motivates you to start your MLOps journey for AI success.