{
"title": "There is no AI salary boom. There are two different booms.",
"subtitle": "Or: the quietly forming two-tier tech workforce in India",
"body": "A few weeks ago, a platform engineer at a services firm sent me a message. His friend had just switched to a pure AI role at a GCC, got a 40 percent hike, and was now in the middle of a second offer bidding war. My candidate was doing the same job he was doing last year, same tech stack, same kind of work. He got an eight percent appraisal. He asked: am I doing AI wrong?\n\nHe is not doing AI wrong. He is doing AI support work, which the market treats differently from AI-core work. The numbers people have shared with me tell a clear story, not a happy one. AI-core roles, the ones building models, fine-tuning them, deploying them, earn year-on-year salary growth of 16 to 21 percent. AI-support roles, the ones working on data pipelines, integration, testing the infrastructure around those models, are growing at seven to eight percent. That is the difference between doubling your salary in four years and doubling it in ten.\n\nI hear from recruiters and hiring managers too, and none of this surprises them. The demand for people who can actually train a model, not just call an API, is genuine. It is not inflated. There are more roles than people who can fill them. The support layer is also hiring, but there is a steady supply of engineers with Python and some data work on their resume. Supply met demand. Growth normalised. Eight percent is not a bad number by historical IT services standards. It is just not the number people hear in headlines.\n\nThe problem is that both these people get grouped under \u201cAI hiring\u201d in the press and in company announcements. A company says it is hiring two hundred AI engineers, and most people think two hundred people will get twenty percent hikes. But a hundred and sixty of those positions might be support. The forty core roles pay well. The rest get the usual market rate. The two tiers are real, but the dividing line is invisible until you are inside one of them.\n\nA composite from three conversations this month: a data engineer at a Bangalore product company told me his team was renamed from \u201cData Engineering\u201d to \u201cAI Data Infrastructure\u201d during the quarterly reorg. No new responsibilities, no new budget for hiring, same manager. The title change was for investor and client optics. He asked if he should list it as AI experience when he switches. I said: depends on what you actually did. If your work did not change, neither will your offer.\n\nThe real divide is not skills either, not entirely. A platform engineer who learns model deployment and inference optimisation can cross the gap in six months. But the market has already priced the difference in. The person doing the deployment is support. The person deciding what to deploy and how to tune it is core. Until you cross that decision boundary, you are in the slower lane.\n\nI told the candidate who messaged me that his friend\u2019s 40 percent hike was real, but it was not a signal about the whole market. It was a signal about one specific narrow slice of it. If he wanted that slice, he would need to shift what he works on, not just where he works. That felt honest and unhelpful both at once, which is often how the truth lands.\n\nThe two tiers are not going to merge. They are going to diverge further. Companies will keep using the word AI for everything because it sounds good. The person writing the integration test and the person designing the reward model both work on AI. They just do not get paid the same.\n\nYou can find your spot at itszia.ai or just figure out which lane you are in. Both are fine. One of them just pays more."