The rise of Large Language Models has made it hard to predict how AI will reshape the world, and the job market is already changing. Rather than listing data science job titles, this piece looks at what changed in the field, why new roles are emerging, and how professionals can prepare for them.
Before the transformer era, data science was largely split in two. Data Scientists sat closer to the product side, focusing on statistics, experimentation, data analysis and insight extraction, with Python and SQL as core tools. Machine Learning Engineers worked nearer the model itself, handling training pipelines, optimization, neural networks, performance and deployment, typically with libraries such as TensorFlow and PyTorch. Data Analysts and Data Engineers had clearer boundaries.
The Applied Scientist title, popularized by companies such as Amazon and Microsoft, filled a middle ground. Applied Scientists experimented with new models and algorithms but also had to ensure their code could operate inside a production system. The lines between roles held because prototyping demanded weeks of manual coding, testing and debugging, with no coding assistants to help.
Everything shifted with the 2017 Google paper "Attention Is All You Need," which introduced the transformer architecture underlying LLMs such as GPT. People in the field began paying serious attention after GPT-3, but GPT-4 was the real turning point, as companies began to rethink how models were built and deployed. That shift is what has given rise to the newer roles examined in the rest of the article.
Source: Towards Data Science · Summarized by HeadlinesBriefing