What do AI employers in South Asia ask for? We read the job boards of five countries.

We read the live job ads on one local job board in each of Sri Lanka, India, Pakistan, Bangladesh and Nepal in October 2026. The short version: outside India, a job titled as AI work is rare. It was 20 ads out of 9,386 on the Sri Lankan, Pakistani, Bangladeshi and Nepali boards, about one in five hundred. But AI is already turning up inside ordinary software jobs. Between 6% and 9% of software, data, QA and IT ads in Sri Lanka, Pakistan and Bangladesh ask the candidate to use or build with AI, and in India it is about one in five. In India, where AI roles are plentiful, the ads ask for Python, LLMs, agents, RAG and evals far more than for model training. They want five years' experience, and most of all they want people who have shipped to production.

This is our own count, not a quoted survey. The method, the limits and the counts are all below, and the counts file is published with the article.

Which job boards did we read, and how?

One board per country, chosen because it allows automated reading in its robots.txt. That ruled out two of the biggest: Naukri in India blocks AI agents by name, and Bdjobs in Bangladesh blocks every crawler. We respected both.

For Sri Lanka, Pakistan, Bangladesh and Nepal we listed every live ad on the board and classified it by its title. For India the board is far larger and IT-only, so we ran twelve AI searches (machine learning, generative AI, LLM, data scientist, MLOps, AI engineer and others) and ten software searches, and read every result. The ads were collected on 18 October 2026, one request a second.

An AI role here means the title names AI, machine learning, LLM, NLP, computer vision, generative AI, MLOps or data science work. A sales, marketing, recruiting, content-creation, annotation or teaching job with "AI" in the title does not count. Pakistan needed this rule most: one recruitment firm called Recruit AI posts dozens of sales and HR jobs under its own name. Other tech means software, data, QA, cloud, IT and similar titles, minus the AI roles.

How many AI jobs are there on local boards?

Very few, outside India.

Those 20 ads came from a small number of employers. In Sri Lanka, Dialog Axiata posted four of the eight: an ML engineer, a senior lead ML engineer, a data scientist and an AI architecture and governance manager. PBA Systems posted two technical-lead roles, and AIClub and Omobio one each, the last a Head of AI. In Pakistan they were AI and LLM engineering roles at software firms (Bluegeko, Anwarson, Skilldockhq and two at Dissolve Labs), plus a data-science unit head at the National Bank of Pakistan. Two more Pakistani AI ads did not load. In Bangladesh: an MLOps engineer, a senior data scientist, an AI internship and an AI research and quality analyst.

This does not mean nobody in Colombo, Lahore or Dhaka is hired to build AI. Much of that hiring happens on LinkedIn, through company career pages, or remotely for employers abroad, and we did not sample those. It does mean that a local job seeker scanning the main local board will see almost no ads titled "AI engineer".

Are ordinary software jobs asking for AI skills?

Yes, and this is the bigger change. We counted the other tech ads that mention AI at all. Then we read each mention, because many are just the employer describing itself ("an AI-powered platform") rather than asking anything of the candidate.

What they ask for is practical. A .NET developer role in Pakistan asks for hands-on use of AI coding tools such as Cursor. A Bangladeshi backend role says to use AI and agentic tools in your daily workflow. A Sri Lankan QA role wants AI-powered testing tools. A Nepali full-stack role asks you to review AI-assisted code. In India the asks run from "familiarity with OpenAI, Gemini and Claude" to building the interface for an AI platform.

So the AI skill a South Asian employer is most likely to want from you right now is not a separate AI job. It is a software job done with AI.

What do AI job ads in India ask for?

Here the sample is large: 1,289 AI-role ads on hirist.tech, posted between 21 July and 17 October 2026. Each skill was counted once per ad, so every figure is the share of the 1,289 that mention it.

The LLM stack leads, but classic ML has not gone. PyTorch and TensorFlow each appear in about a third of ads, far more often than in our September count of mostly US and European ads, where PyTorch was in 5%. Part of that gap is length: hirist ads are long (the median is about 2,600 characters), and longer ads name more tools. The order is the more reliable signal, and in India the order is: Python, then LLMs, agents, RAG and evals, then the frameworks.

On cloud, Azure (50%) edges AWS (48%), with GCP at 36%, and MLOps tooling is in 46% of ads. The non-tool words are the loudest: production or shipped in 72%, system design or architecture in 67%, stakeholders or cross-functional in 55%, communication in 51%.

How much experience do Indian AI ads want, and what do they pay?

Five years is the median minimum. Every hirist ad states a range: 61% ask for five years or more, and only 181 of 1,289 (14%) accept two years or fewer. Bangalore is named in 578 ads, then Hyderabad (247), Pune (181), Gurugram (144) and Chennai (133). 112 say anywhere in India. Hybrid is mentioned in 18% and remote in 9%.

A degree is mentioned in 36% of ads, using the same wording test as our global count, where it was 12%. A PhD appears in 6%, a certification in 9%, and a portfolio, GitHub profile or open-source work in 25%.

Pay is rarely shown: 96 of the 1,289 ads (7%) give a range. In those, the median midpoint is ₹32.5 lakh a year and the middle half runs from ₹25 to ₹45 lakh. Ads that show pay skew senior, so treat this as the going rate for experienced AI engineers, not a starting salary.

What should a South Asian learner take from this?

Three things, each from the counts above.

1. Learn to build software with AI, not only to build AI. Outside India, the ads that ask for AI skills are mostly ordinary developer, QA and data jobs that expect AI tools and AI features. That is the larger market today.
2. If you want an AI-titled role, the stack is Python, LLMs, RAG, agents and evals, on a cloud. That holds across all five countries. The Indian ads also keep PyTorch and statistics in the picture, especially for data-science titles.
3. Shipped work beats a list of tools. Production experience is the most common requirement in Indian AI ads, ahead of every tool except Python. A deployed project you can show answers it more directly than a certificate does.

If you are choosing where to learn, our country pages for Sri Lanka, India, Pakistan, Bangladesh and Nepal cover local salaries and options.

What are the limits of this count?

Five, stated plainly:

• One board per country. LinkedIn and company career pages were not sampled, and in Sri Lanka in particular much tech hiring happens there. Naukri and Bdjobs, the largest boards in India and Bangladesh, were excluded because they block automated reading.
• India was searched, not counted in full. hirist.tech is IT-only, so there is no share-of-board figure for India, and the 1,289 ads are what the AI searches returned.
• Small numbers outside India. 20 AI roles across four countries is enough to say AI titles are rare. It is too few to describe what they ask for, so we did not.
• Titles decide the groups. A "Software Engineer" building an AI product counts as other tech, not as an AI role.
• One snapshot. These are the ads live in mid-October 2026. We plan to repeat the count each quarter and publish a new dated file alongside this one, not overwrite it.