How Leonid took courses on bots and neural network
Not long ago, I decided to take a course on Telegram bot development. In general, I was already familiar with the topic and had even built a framework for integrating synchronous bots with Django.
Sporting a very original name:
Django Telegram Framework 😄
But that is a story for another day.
I noticed that bot development gigs pop up quite frequently on freelance platforms, so I decided to take a course at OTUS: "Conversational Bots and Voice Assistants".
As far as bots go, I didn't really learn anything new 🙂 The main library used there was aiogram.
On the bright side, I learned how to work with neural networks (LLMs). Surprisingly, the topics covering tools resonated with me the most.
In short, it is all about teaching an LLM to utilize various instruments — like making API calls or executing custom-written functions.
You can achieve this either directly by passing a set of functions in the code, or by providing the model with a link to an MCP server that describes the functions and handles their execution.
The MCP topic caught my interest so much that I based my final capstone project on it, just to test out how it handles in practice.
And as it turned out, it is a pretty incredible piece of tech 🚀
Oh, right, there was also RAG — a rather complex topic indeed 🙂
The takeaway and conclusion:
The course is officially completed, and I have got a certificate to show for it 🏆
Ready to be hung on the wall 😄
I initially thought it would be dry and not that exciting, but it turned out to be a blast and highly engaging.