The ML Block

Inspiration

Airtable has disrupted Software development, drastically reducing the time to build products. Also, the flexibility of the platform has enabled developers like us to extend it's functionality further.

We thought, what if people like Data scientists, or for that matter anyone, use their data to gain valuable insights?

What it does

The ML block allows you to train to Machine Learning models within Airtable itself, without knowing about the technicalities! Just use the GUI to select the appropriate model, dataset, and then use the model to make predictions!

How I built it

• Created the interface using Airtable UI components
• Used Axios for making HTTP requests.
• Created a flask server which trains models as per user's requests, and pickles(saves) them to the disk.

Challenges I ran into

• This was my first acquaintance with React. Learning about state management, hooks, asynchronous way of programming were a bit overwhelming at first.
• Getting acquainted with a new platform and paradigm of software development ( using reusable blocks ) was a bit challenging. But the documentation helped immensely!

Accomplishments that I'm proud of

• Building a functional and useful product!

What I learned

• React - and useful concepts like State management, asynchronous programming
• Good practices

What's next for The ideal Teacher

• Give more customization option to users with technical knowledge - hyperparameter tuning ( number of epochs, train test split ratio)
• Model versioning

Built With

• axios
• flask
• javascript
• python
• react
• sklearn