how to stay relevant as a Data Scientist

As Alexey Grigorev stated in one of his lessons in LLM Zoomcamp: "To be honest these days coding feels like cheating with all LLMs. You just give it an error and then keep your fingers crossed and hope for the best that it works". By the way, If you're interested in building LLM-based applications (like RAG) and don't have a senior collegue to learn from them, checkout the LLM Zoomcamp created by DataTalksClub. I’ve found their hands-on coding videos and practical approach incredibly valuable. It's fascinating to see Alexey using Claude.ai to generate initial code and then refine it into something functional 🧙‍♂️

While debates about AI hype and the possibility of an AI winter continue, one thing is clear: LLMs are significantly benefiting code developers, especially data scientists, by saving time and streamlining the transition from "ML model in Jupyter notebook" to "deployable application".

You just need to define the system (e.g streamlit app with docker-container with elastic for data-storage and Postgres to log parameters, etc), ask LLM to generate the code and almost immediately you will get a bunch of files. Though it might not work perfectly right away, it’s a huge leap forward compared to what could be achieved manually in minutes.

This shift is changing the role of everyone involved in the development process. Not long ago, we were coding in assembly languages tied closely to the hardware. 😱 Today, many of us can program at a higher level of abstraction, focusing less on hardware specifics and more on functionality.

With LLMs advancing coding to a new level of abstraction, we now should think in terms of systems and their interactions rather than individual scripts. It's true that LLM still can't write bug-free code, but eventually we will get there. Designing systems and processes will remain a human task for the foreseeable future. So, if you're wondering how to stay relevant as a Data Scientist, it's time to start building your skills in ML/AI Engineering — such as cloud computing, MLOps, and ML system design. These skills are increasingly essential for senior roles and are no longer just "nice to have" but rather a "must."