Adopting MLOps Methodology in Zeals
Introduction
Hello There, this is Rituraj and I work as the Data and ML Engineer at Zeals.
If you want to learn about how Zeals is Adopting MLOps Methodology you have come to the right place.
Let us begin!
Background Story
We have data scientists and operations professionals on our team, and while they do their best to collaborate, they each have their own areas of expertise.
This could lead to miscommunications and misunderstandings. The data scientists can interpret the data, but they likely don’t have the background to manage business operations. Likewise, the business team can make their side of things work, but they don’t fully know how to interpret and implement data.
Here is where machine learning operations (MLOps) and GCP Vertex AI comes in.
What is MLOps?
MLOps (machine learning operations) is a set of best practices for improving communication and collaboration between the data science and operations team that aims to make developing and maintaining production machine learning seamless and efficient.
It is an extension of what we know as DevOps. DevOps is the process of organizing cooperation between everyone involved in the design and building of big data. This process has been around for a while, but MLOps is still in its youth.
The data science community generally agrees that it’s an umbrella term for best practices and guiding principles around machine learning — not a single technical solution.
The Four guiding Pillars of MLops
- Machine learning should be collaborative.
- Machine learning should be reproducible.
- Machine learning should be continuous.
- Machine learning should be tested & monitored.
Why MLOps Matters?
The goal of MLOps is to reduce technical friction to get the model from an idea into production in the shortest possible time to market with as little risk as possible.
Traditionally, machine learning has been approached from a perspective of individual scientific experiments predominantly carried out in isolation by data scientists. However, as machine learning models become part of real-world solutions and critical to business, we will have to shift our perspective, not to depreciate scientific principles but to make them more easily accessible, reproducible, collaborative, and most importantly to increase the speed at which machine learning capabilities can be released.
MLOps creates a lifecycle and a set of practices that apply to the development of machine learning systems. This includes research, development, operations, and implementation
The process of brainstorming, developing, and implementing machine learning is extensive. Having a set of duplicatable processes to guide each project helps in many ways.
Why did we need to adopt MLOps?
Creating an ML model that can predict what you want it to predict from the data you have fed is easy. However, creating an ML model that is reliable, fast, accurate, and can be used by a large number of users is difficult.
The necessity of MLOps can be summarized as follows:
- ML models rely on a huge amount of data, difficult for a single person to keep track of.
- Difficult to keep track of parameters we tweak in ML models. Small changes can lead to enormous differences in the results.
- We have to keep track of the features the model works with, feature engineering is a separate task that contributes largely to model accuracy.
- Monitoring an ML model isn’t like monitoring a deployed software or web app.
- Debugging an ML model is an extremely complicated art.
- Models rely on real-world data for predicting, as real-world data changes, so should the model. This means we have to keep track of new data changes and make sure the model learns accordingly.
How did we use MLOps Methodology?
We also had the same expression as above when we were are trying to convert our understanding of all the above MLOps learning into practice.
Our team explored and compared the pros and cons of different tools and services across the open-source tools and cloud-based managed services.
We not only emphasized choosing a unified MLOps Technology but also focused on having a platform where every team member in the organization whether he is an application developer, data analyst, or even business user gets the opportunity to explore the capabilities and benefits of augmenting their workflow or products with ML even if they don’t have the expertise required to create ML models.
The managed service of different platforms like GCP, AWS, and Azure were very fascinating as it had made the implementation of the MLOps methodology way much easier by removing the overhead complexity of
- Infrastructure Setup
- Management
- Maintenance
- Scalability
- Security
- Integration
After comparing different existing Cloud-based Services tools and taking into consideration of our existing Cloud applications we had decided to adopt the GCP Vertex AI as a one-stop platform for MLOps.
What is Vertex AI ?
Vertex AI brings together the Google Cloud services for building ML under one, unified UI and API. In Vertex AI, you can now easily train and compare models using AutoML or custom code training and all your models are stored in one central model repository. These models can now be deployed to the same endpoints on Vertex AI.
Vertex AI integrates the ML offerings across Google Cloud into a seamless development experience. Previously, models trained with AutoML and custom models were accessible via separate services. The new offering combines both into a single API, along with other new products. You can also migrate existing projects to Vertex AI.
In addition to model training and deployment services, Vertex AI also includes a variety of MLOps products, including Vertex Pipelines (the focus of this lab), Model Monitoring, Feature Store, and more. You can see all Vertex AI product offerings in the diagram below.
Conclusion
Phew! That was quite a lot of Theoretical Knowledge Adventure. Give yourself a pat on your back. 👋
We covered
- What is MLOps?
- Why does it matter?
- Why did Zeals need to adopt it?
- Why did we adopt the GCP Vertex AI Platform for MLOps and its overview?
We will try to bring in practical use cases in our upcoming blogs to show you how our company is adopting and utilizing MLOps methodology with GCP Vertex AI.
Stay tuned for that next piece of blog! Thank you for reading, and stay safe!
Reference:
- Comparison of Cloud-based MLOps platforms by Vineet Jaiswal — Awesome blog recently published with the same idea and comparison stats as Zeals. (It came out when I was drafting this blog post. He has done justice to it and so we did not want to recreate it and used the same comparison table)
- New To ML: Learning Path On Vertex AI
- A Gentle Introduction to MLOps by Yashaswi Nayak
