Only 16 junior AI roles in 2,721 job postings

Every week someone asks me how to get their first AI job.

So last month I actually sat down and went through 2,721 real AI job postings from one of India's biggest tech job portals. Skills are tagged by the platform itself, so no guesswork on my end.

Three things from that data changed how I think about AI careers.

 

1. The market is not built for freshers

Out of 2,721 postings, only 16 were junior or intern roles. Sixteen.

1,555  —  mid-level roles

1,040  —  senior roles

14  —  junior roles

2  —  intern roles

94% of AI jobs want 2+ years of experience. This is not great news if you are starting out, but it is the reality. Better to know it now than after 6 months of applying.

 

2. RAG is winning. Fine-tuning is not (yet)

I genuinely expected fine-tuning and LoRA to show up everywhere given how much content exists around it online. Instead:

486  —  jobs mention RAG

384  —  jobs mention Agentic AI

303  —  jobs mention LangChain

11  —  jobs mention Fine-tuning + LoRA combined

0  —  jobs mention CrewAI

Fine-tuning is a real skill and LoRA is genuinely useful, just not what companies are hiring for right now. Learn RAG first. Fine-tuning can come later.

 

3. The most underrated skill in the whole dataset? ETL.

Not Agentic AI. Not LangChain. ETL — the boring data engineering skill nobody makes content about.

380  —  jobs mention Data Modeling — more than LangChain

334  —  jobs mention ETL — more than Kubernetes (49)

279  —  jobs mention Data Pipeline

Fundamentals still win. Companies want people who can move and shape data properly, not just people who know the latest framework.

 

If you're a fresher: Python and SQL first, then RAG, then one real deployed project. Skip the fine-tuning rabbit hole for now.

If you're experienced: add the GenAI layer to what you already know. And if you're a mid-level GenAI engineer aiming for senior — MLOps is your gap. It shows up in 0% of mid-level job posts and 14% of senior ones.