Photo by Tran Mau Tri Tam on Unsplash
Background Story:
Recently, I participated in a hackathon event. The hackathon event is a global event from Hackmakers called the #BuildwithAI2021. There were participants at all levels — students, freshers, experienced and they had mentors/lead mentors who support/guide the participants.
The event is similar to all the other hackathon events except that I have participated in a “team captain” this time. It is a big role in fact this is my second hackathon event that I participated in.
We were given a problem statement and right off the bat started with ideation, and solution building. we had mentored to guide us and all was going well but the tough came later.
The solution should be in a form of an app or a website and none of us were experts in front-end skills. That's when I realized the fact that how important is a model deployment phase. You can build whatever ML/DL models no doubt in that however what is the use of it if it is not been used for the latter purpose or deployed for future.
Model deployment - why is it important and do we really care:
It is obvious!! Model deployment is the only way to turn your model into a practical solution to make business decisions.
Yes! we care, why? because all of us want to make an impact and what better way to do it through practical means. The only way we can make sure the model is been used is through model deployment
Model deployment is the method by which you integrate a machine learning model into an existing production environment in order to start using it to make practical business decisions based on data. - Microsoft
Photo from Microsoft Azure ML
Different ways to deploy a ML model:
There are several ways to deploy an ML model, some of them are below:
Using the existing Cloud Platforms:
Heroku is a cloud Platform as a Service that helps developers quickly deploy, manage, and scale moderns applications without infrastructure headaches.
Google Cloud Platform (GCP) is a platform offered by Google that provides a series of cloud computing services such as Compute, Storage and Database, Artificial Intelligence(AI) / Machine Learning(ML), networking, Big Data, and Identity and Security.
Microsoft Azure Functions is a serverless cloud service provided by Microsoft Azure as a Functions-as-a-service (FaaS).
Amazon Web Services Lambda is a serverless computing service provided by Amazon as part of Amazon Web Services.
Using Softwares:
KNIME software is widely used for many areas from EDA to creating ML/DL models. however, one thing that it can serve is to automate the entire model workflow and use it to be deployed in production.
Flask is a web framework. IT can be used to build a highly customized solution from the ground up.
Streamlit is a data dashboarding tool. Streamlit is an all-in-one tool that encompasses web serving as well as data analysis.
Dash lets developers customize the front-end by writing HTML and CSS. It can be used to build more customized data dashboards for non-technical users.
Having the skill to deploy is not mandatory but can be very well learned by data scientists in need of the hour.
As there are many ways to deploy, one can choose which ways they prefer to deploy based on their needs/purpose.
Resources:
If you like this post feel free to subscribe to this weekly newsletter for weekly updates on data science related topics.
Comments