How to Get Away with No-code
and a list of some no-code / low-code tools to build data MVPs ultra fast
For my first sprint as a Data Scientist in an IoT company, I had to scrape a bunch of images from the internet which would later be used to train a neural network for image classification. My first thought is using either fastai or BeautifulSoup or Selenium + ChromeDriver. But something else hit me.
The first principle of scrum: Pick what’s fastest. Deploy. Get feedback. Iterate. Again, it’s a way to Hello World your way into building things that work and are useful to your client. Well, the fastest way to get into this scraping, I thought, is a no-code solution. And there I was, getting through the first week of my neural network project by using very little code, rather just
- Looking for publicly available datasets that contain the images needed and requesting for some datasets by sending an email to academics/researchers, living across the world, asking for access to those datasets
- Chrome Web Extensions (like, Fatkun Batch Download)
- ImageNet and Free Download Manager
- Some public Github repositories that could scrape google and bing for photos
With this minimal code solution, I was able to batch download approximately 180k photos from the web (and from those sent by the researchers, thanks to all of you!). All of this was working UNTIL…
…the second wave of sprints came in. As I was already moving on to productizing / deployment of the neural network I just trained, I was given 2 more projects to work on, one of which, again, involves web scraping.
This time, I have no choice but to do it on BeautifulSoup or Selenium. Ha!
So, this forced me into a refresher course in HTML and CSS.
I am so far finding this Traversy Media YouTube series very helpful as someone who took HTML and CSS classes 3 years ago and have been using Streamlit or Flask templates in Github to create web apps ever since.
Here’s the Hello World I ran on localhost
Here are some more no-code / low-code tools I encountered recently, for rapid prototyping:
- figma — this one I actually used for a long time. I used this for wireframing a plant production and chemical inventory app for a hydroponics farm. User feedback was positive on the wireframe and I attribute the good design from all the useful templates and design elements readily available on figma.
I will chronicle my Data Science journey here so together, we could start to “be comfortable with being uncomfortable”. That’s what it takes to become a good software developer anyway. Just like with anything else in life and in any other industry, when you encounter a steep learning curve, it just means you are about to learn a lot and grow a lot.
Here’s to all the growing pains we’re about to get!