Python is a great industry standard language that can easily handle both machine learning and web development tasks. Our dev team is very familiar with the language and has used it in various web and Machine learning projects. Python has many versatile ML specific libraries that include TensorFlow, Pytorch, Pycaret, and Keras. It also has packages for data manipulations and visualization like Numpy, Pandas, and Matplotlib. Since our software requires machine learning algorithms, big data processing and a backend server, Python seemed like the way to go.
Our team decided to go distributed databases (NoSQL) over a relational database (SQL) because of the NoSQL dynamic schemas for unstructured data. We are using MongoDB as our NoSQL database due to its simplicity, schema less documentation, deep/fast querying ability, user data management, big data, JSON style documents, and great scaling out. We also chose MongoDB due to its horizontal scaling as a NoSQL database.
Since we are using python as our backend programming language, we decided to use Flask as our web framework. Flask is a micro and lightweight web framework that provides the required functionality to efficiently develop our web server. Flask has a great community with many online resources and provides more flexibility in terms of customization when compared to other frameworks like Django. While Django is great for large scale applications, it does not work well with NoSQL databases.
For our front end framework, we decided to go with React due to its component based structures, flexibility, scalability, and high performance. React has a strong community and is trusted by top companies such as Facebook, Netflix, and Paypal. We can also easily transition our react app to a react native or electron app. We will also be using material-ui framework alongside react for that crisp google material design!
Node.js will be used for development purposes for the front end only. Once we deploy for production, the react frontend will be served from the flask web server and will not require Node.js. This separates the frontend and backend during development, making it easier to work with.