Learn Fast and Build Things: How Julia Kryuchkova Went From No-Code to Writing Python Apps

_{Julia generated this image using ChatGPT.}

Julia Kryuchkova is a program manager at First Line Software, a DeepLearning.AI student, and a beginning programmer. After bouncing off a few other introductory Python courses, she found success with our AI Python for Beginners course series. In this story, she describes how she used chatbot help and her own know-how to build her first Python programs and how she plans to use these new skills in her life and career.

Recently, I completed the short course AI Python for Beginners by DeepLearning.AI, taught by Andrew Ng. I am not a programmer, so I was thrilled when immediately after completing the first course in the sequence — which took about an hour — I managed to create my first program in Python. Then I created another one. And one more. The next day, I even managed to automate one of my routine tasks, which would have required quite a lot of manual work… if it weren’t for the techniques I learned from the course.

My projects

1. Python Temperature Converter

My first program, inspired by one of the course’s projects, was a calculator that converts temperatures from Celsius to Fahrenheit. I thought it would be nice to be able to tell my overseas colleagues what the temperature is right now in a way they would understand. ChatGPT suggested the code for the program, but the input field was complicated and the converter only worked one way (it could convert Celsius to Fahrenheit but not vice-versa). So I asked the chatbot to simplify the input and improve the user experience. ChatGPT came up with a genius solution: depending on whether you add a “C” or an “F” to the temperature measurement, it converts Celsius to Fahrenheit or back. For example, “20C” converts to Fahrenheit, and “10F” to Celsius.

An early version of Julia's temperature converter. It asks for temperature in Celsius or Fahrenheit. — _{Julia’s Python temperature converter’s input field.}

Julia's Python converter shows some of the code that processes the output. — _{Julia’s Python temperature converter showing the output.}

2. The “Time Machine”

My next idea was to calculate the current time in all of our team members’ time zones. Here it took me much longer to find the format and the output I needed. At first, when I was trying to add locations, the program needed me to add exact longitude and latitude coordinates along with the exact time zone. Another detail that I really wanted was a nice big font to help accent the most important information. But after about an hour of working with ChatGPT, I got what I wanted: easily added locations and nice bold output.

Julia's time zones clock shows some of the code in process and converts the current time to the time in Budva, London, Boston, and Los Angeles. — _{Julia’s Python time zones clock, now with easy-editing locations and nice large output.}

3. The Shabbat

When I showed this “Time Machine” program to my friend, she asked if I could create a program that calculates the start and end of Shabbat. It required some more “ChatGPT thinking” and installing some libraries to calculate sunset times. But it worked!

4. Website Summaries

The next day, I thought about automating one of my most time-consuming tasks at work: creating summaries for our new website pages. Our website is still under development, and the content is password-protected to prevent accidental access by the public. There are no secrets; it’s just a work in progress.

I understood that building a summarizer was complex, so I broke it down into parts. First, I asked ChatGPT to create a simple program that would summarize a publicly available webpage. After this was done, I tried to use the Gemini AI assistant in Google Colab Notebook to add new functionality for summaries of password-protected pages. Gemini refused, probably due to some rules preventing it from assisting with such tasks. So I turned back to ChatGPT and explained the situation. ChatGPT’s initial solution didn’t work either; it couldn’t get past the password protection. Eventually, I copied part of the HTML code from the login page into ChatGPT. That did the trick! My next version of the program worked. I could add my password into the system and access the pages I needed to summarize.

Next, I expanded the program. I added the ability to include multiple links and save the results in the format I wanted — a CSV file (a common format for storing simple spreadsheets and other tabular data). After that, when I ran my program, it completed the task in seconds.

However, I soon discovered that I wasn’t entirely happy with the quality of the summaries themselves. ChatGPT explained that this might be due to the built-in-Python library I used for my summaries.

Now that I had a program to streamline my workflow, my next task became trying to improve the summaries. I asked a friend for help, and ChatGPT was also able to make some useful suggestions. Eventually, I connected my Python script to OpenAI’s GPT-4o using my API key. I also modified the script to make sure that it would only summarize the content I need, not the menu or service info that is repeated on all the pages. After this, my website summaries were much better!

This codeblock shows part of Julia's Python script for creating webpage summaries. — _{Part of Julia’s Python script for creating webpage summaries.}

Conclusion

I would like to say thank you to Andrew Ng and the other course authors. They truly inspired me to start creating new stuff, to ask questions and get answers.

The course and these projects also helped me understand that developers spend a lot of their time on finding and fixing problems – debugging, fine-tuning, testing – after the basic functionality is already working. This can take much longer than just writing or generating code, or even getting the program to work in a limited way.

After I built these projects with what I learned in Basics of AI Python Coding, I was ready to complete the other three courses in the series: Automating Tasks With Python, Working With Your Own Data and Documents In Python, and Extending Python With Packages and APIs.

These courses bolstered my confidence. Now I feel much more secure in my knowledge of Python: I know that I can do some programming when needed with the aid of generative AI models. I started my journey with DeepLearning.AI by taking ChatGPT Prompt Engineering for Developers, and now I also know that I can handle more sophisticated prompt engineering tasks that require basic knowledge of Python. My non-technical background won’t stop me from taking on this role, which is really important for moving forward.

Finally, I feel more confident in my ability to learn new skills. Even though I know there’s a huge difference between me and an experienced programmer, I am happy that I can now automate my tasks with Python alone, without relying on Zapier or other third-party tools.

If you need more complex programming, you should come to real programmers like the ones in my company (First Line Software). But for small automation tasks like the ones mentioned above, I highly recommend the AI Python for Beginners course series.

Two last tips: I would advise anyone using AI to help them with beginning programming to take this course sequence to learn some basics about how coding works. Otherwise, the code you generate will look like total magic that you have no way to understand. If you don’t understand the basic elements of the code, you might get stuck and be unable to get what you want from the system. I was able to extend my work beyond the course’s beginner projects largely because I now knew Python well enough to read and modify new code generated by ChatGPT.

It’s also a great idea to start with some small and simple programs. If you can’t invent such tasks on your own, try asking ChatGPT: it’s great for generating ideas. 🙂