Key insight 6 of 10
Data Science & ML has shifted toward LLM application engineering
One of ten findings from the analysis of 4,022 remote positions, with the supporting numbers recomputed from the underlying listings.
The finding
In the 422 DS&ML roles, RAG (25) nearly matches Machine Learning (29). LLM-related tags also rank highly: LLM (20), LLMs (17), Generative AI (17), LangChain (18) and Vector Databases (16). Across all roles, RAG appears in 38 positions, LangChain in 26, Vector Databases in 22, LangGraph in 19 and MCP in 13. Classic frameworks are smaller: PyTorch has 18, TensorFlow 13 and scikit-learn 11.
From the published analysis. Counts use the original skill names, before naming variants were merged.
Top skills in Data Science & ML
Merged naming variants; share of 422 positions
LLM tooling and classic ML across all roles
Positions naming each skill
Unique positions naming any LLM tool listed here: 129. Naming any classic ML skill listed here: 70.
What it means for job seekers
- LLM application work (retrieval, orchestration, vector search, evaluation) now rivals classic machine learning in data science and ML listings.
- Python remains the foundation for both branches.
- The engineering skills around the model, such as evaluation, cost control and deployment, matter as much as the framework you choose.