As artificial intelligence reshapes every aspect of IT, job titles in the data/AI space have multiplied and confused both candidates and hiring managers. At Inuits, our Talent-as-a-Service (TaaS) approach relies on having clear, realistic definitions of what these roles do today, not just what the hype suggests.
Photo Credit: Precessor, via flickr.com
While companies are still experimenting with AI-powered roles, some solid patterns have emerged. Let’s kick off with the most prominent. Back in mid-2023, an influential Latent.space blogpost counted about 10* as many ML Engineers vs AI Engineers on Indeed, but predicted an inversion of that ratio in 5 years. Fast-forward two years, the reality has become more nuanced with 18k jobs for ML Engineer and about 11k for AI Engineer.
Data from Curioz shows that Poland’s AI job market has shifted from steady to sprinting. In late 2024, around 2.4% of all jobs were AI-focused, with only small monthly changes, reflecting a niche but stable demand. Early 2025 brought a sharp hiring surge, jumping from 2.77% in February to 4.62% in April, and reaching 5.56% by July, more than double the share from last September. The fastest growth happened between February and April, likely fueled by fresh corporate budgets, major AI project launches, and possibly government or EU-backed funding programs. This rapid rise suggests that AI adoption in Poland is moving from experimentation to large-scale implementation, with companies actively building dedicated AI teams. AI roles are no longer niche, they are becoming a core part of the market, and the momentum indicates that their share could continue climbing through the rest of 2025.
Meanwhile, entirely new titles are popping up, we see the emergence of Prompt Engineers, Generative AI Engineers or even GPT Engineers.
Who are they, and why?
Fig 1: Percentage share of AI-focused Job Roles in Poland
Our view, reinforced by analysis from Curioz.io, is that almost every AI and data job in 2025 can be mapped to six high-level categories. Everything else is either a buzzword, a niche specialization, or an overlapping title.
The Six Core AI/Data Job Categories
-
Data Analyst
Core focus: Interpret and visualize data to answer business questions.
Skills: SQL, BI tools (Tableau, Power BI, Looker), data storytelling.
Tools: Excel/Google Sheets, visualization libraries, SQL-based warehouses.
Reality check: In AI contexts, analysts increasingly use AI-powered BI tools but still focus on insight delivery, not building ML models.
-
Data Engineer
Core focus: Design and maintain data pipelines, storage, and ETL processes.
Skills: Python/Scala, distributed processing (Spark, Flink), orchestration (Airflow), cloud data infrastructure.
Tools: AWS/GCP/Azure data services, dbt, Kafka.
Reality check: Without clean, well-structured data pipelines, AI projects stall before they start.
-
Data Scientist
Core focus: Build statistical and machine learning models to extract insights and make predictions.
Skills: Python/R, scikit-learn, exploratory data analysis, statistics, feature engineering.
Tools: Jupyter, Pandas, NumPy, cloud notebooks.
Reality check: Often overlaps with ML Engineer, the big difference is that Data Scientists spend more time exploring and prototyping, less time deploying.
-
ML / MLOps Engineer
Core focus: Productionize ML models. Deploy, monitor, and maintain them at scale.
Skills: MLflow, Kubeflow, container orchestration (Docker/Kubernetes), CI/CD for ML, model monitoring.
Tools: AWS SageMaker, Vertex AI, Azure ML, Triton Inference Server.
Reality check: This is the critical bottleneck; many AI projects die after experimentation because they lack people who can actually deploy and maintain models reliably.
-
ML / AI Researcher
Core focus: Push the boundaries of models and algorithms, from training large language models (LLMs) to experimenting with new architectures.
Skills: Deep learning theory, PyTorch/TensorFlow, CUDA, distributed training, academic research.
Tools: Hugging Face Transformers, custom training pipelines, high-performance compute clusters.
Reality check: Only a tiny fraction of companies train foundational models from scratch. Even most researchers now build on top of open-source LLMs or models from the big players.
-
AI Engineer
Core focus: Integrate and adapt existing AI models (often LLMs) into products; build surrounding infrastructure like retrieval pipelines, embedding systems, and prompts.
Skills: Backend development (Python/Node/Java), vector databases (Pinecone, Weaviate), RAG pipelines, API orchestration, fine-tuning models.
Tools: OpenAI API, Hugging Face Inference API, LangChain/LlamaIndex, cloud functions.
Reality check: Despite the futuristic title, most AI Engineers are not training models. They’re hooking into existing ones and wrapping them in apps. The “heavy” model training usually remains with big AI labs.
What about Prompt Engineers, Generative AI Engineers, GPT Engineers?
They are not separate core roles; they are skill sets or niches within the above categories:
- Prompt Engineering: A core skill for AI Engineers, Data Scientists, and even Analysts using LLMs.
- Generative AI Engineer: Often just an AI Engineer working with text/image/video generation models.
- GPT Engineer: Buzzword variation for someone building with OpenAI APIs.
Photo credit: Elon Musk at TED2022, via TED.com
The Blurred Lines: "AI-Enabled" Roles
In reality, the line between "AI expert" and "AI-enabled" is fuzzy.
Many “AI Engineers” and “AI-Enabled Backend Engineers” do the same thing: connect existing models to applications. The real distinction is not about the title, but about:
- Depth of AI knowledge.
- Ability to adapt/explore models vs. just call APIs.
- Understanding of infrastructure for scaling and monitoring AI in production.
Examples:
- AI-enabled Backend Engineers build APIs that integrate with LLM endpoints, similar to what many AI Engineers do.
- AI-enabled Frontend Developers hook up UI with AI-generated outputs.
- AI-enabled QA/Test Engineers use AI-powered test generation tools, without building AI.
Fig 2: Percentage share AI related terms in open job roles in Poland.
Why This Matters for Talent-as-a-Service
- Most of the market in 2025 is about applying existing AI rather than developing it from scratch.
- Hyper-specialized roles are rare outside of top research labs.
- MLOps and infrastructure expertise are the most underestimated skill sets, without them, AI projects fail to scale.
Pro tip for Talent-as-a-Service teams:
Start with these six functions as the backbone of your AI/data talent strategy, then layer in “AI-enabled” skills and niche expertise as your projects demand.
This way, your team stays agile, avoids buzzword-confusion, and remains focused on delivering scalable impact.
NOTE: This post is based on research by Inuits.it and Curioz.io, and has been crossposted on both platforms.