Hiring AI talent in India
Hire AI engineers in India without setting up an entity
India has the second-largest pool of AI researchers and inventors in the world. It does not have the second-largest pool of hireable AI engineers, and the difference is where most foreign employers lose money. This page gives you the city-level picture, the real employer cost of an India AI hire after the November 2025 labour codes, and the three legal routes to your first engineer.
Can a foreign company hire an AI engineer in India without a local entity?
Yes. An Employer of Record becomes the legal employer in India while you direct the work. There is no minimum entity setup, no registered office and no share capital requirement. A compliant hire can start in days rather than the three to six months an entity incorporation typically takes.
What does an India AI engineer actually cost an employer?
Salary is almost the entire cost. Employer-side statutory contributions on a senior AI salary come to roughly 2 to 3 percent of CTC, not the 15 to 20 percent many founders budget for. This is because provident fund is administered against a wage ceiling and state insurance stops applying well below an AI engineer's salary.
Where in India should you hire AI engineers?
Bengaluru has the deepest senior pool and the highest price. Hyderabad is the strongest market for engineers trained inside global capability centres, at roughly 20 percent less. Delhi NCR pays a fintech and consumer-internet premium. Pune and Chennai are thinner for specialised machine learning work and are better suited to applied AI and data platform roles.
01 — The map
Where India's AI engineers actually are
India's AI talent is not evenly spread and the gap between the top two cities and the rest is wider for machine learning than for general software engineering. Bengaluru and Hyderabad together hold the large majority of senior ML roles in the country. Treating "India" as a single hiring market is the first mistake most foreign employers make.
| City | 0–2 yrs | 3–5 yrs | 6–10 yrs | What it is strong at |
|---|---|---|---|---|
| Bengaluru | ₹8–14 L | ₹18–30 L | ₹35–58 L | Deepest senior pool. AI-first startups, ML platform teams at consumer unicorns. Also the most competitive: expect counter-offers. |
| Hyderabad | ₹6–11 L | ₹14–25 L | ₹28–48 L | Strongest GCC-trained ML pool. Engineers with production experience inside large global engineering organisations. Also a real pharma and life-sciences AI cluster. |
| Delhi NCR | ₹9–15 L | ₹20–35 L | ₹40–65 L | Fintech and consumer-internet ML. Highest headline numbers of the five, concentrated in a narrower set of employers. |
| Pune | ₹6–11 L | ₹14–24 L | ₹28–46 L | Applied AI, data engineering and enterprise product teams. Thinner for research-grade ML; senior candidates often relocate out. |
| Chennai | ₹6–11 L | ₹13–23 L | ₹26–44 L | Enterprise engineering, SaaS and NLP product work. Lowest attrition of the five, which matters more to total cost than most employers model. |
A caveat worth reading. Any single salary figure for "AI engineer in India" is close to meaningless. The gap between an IT services employer and an AI-first product company for the same job title is routinely three times or more. Specialisation moves it again: engineers with hands-on LLM fine-tuning, retrieval systems or LLMOps experience command a substantial premium over generalist ML engineers. Use the bands above to set a budget range, then price the specific candidate.
02 — The filter
Pool size is not hiring capacity
India's AI talent pool is projected to pass 1.25 million by 2027, up from roughly 600,000 to 650,000 in 2022. That number gets quoted constantly and it is not the number you should plan against. Graduate employability studies put overall employability at around 43 percent, rising only slightly to about 46 percent for AI and machine learning roles specifically. Slightly under half of the people with the credential clear an employer's bar.
This is not a reason to avoid India. It is a reason to budget for screening rather than for volume. In practice, three things separate a hire from a résumé:
- Production, not demos. Ask what broke in production and what they did about it. Anyone can describe a model they trained. Far fewer have owned logging, evaluation, versioning, fallback behaviour and cost control on a live system.
- Evaluation discipline. How does the candidate know their system got better? Teams that cannot answer this ship regressions confidently. It is the single sharpest filter in applied AI hiring right now.
- Verifiable public work. Substantive open-source contributions and named conference authorship are higher-signal than certifications, which are close to noise at senior levels.
Screening aside, the mechanics of the hire itself are the same as for any other role. Our guide to hiring employees in India covers offer letters, notice periods, background checks and the documents you need before day one.
03 — The cost
What an India AI engineer costs you, fully landed
Most cost pages stop at salary. The number you actually need is salary plus employer statutory contributions plus the cost of employing them legally. Set the two controls below.
Total annual cost to you
—
—
| Employee CTC midpoint of the city band | — |
|---|---|
| Employer provident fund 12% of wages, administered at the ₹15,000 ceiling | — |
| EDLI and EPF administration also ceiling-linked | — |
| Employees' State Insurance applies only up to ₹21,000 gross per month | — |
| Gratuity accrual 4.81% of wages, provisioned monthly | — |
| Group medical and accident cover Saileor standard policy | — |
| Employer statutory and benefits, total | — |
| Saileor EOR fee flat, does not scale with salary | — |
—
Hiring for something other than an AI role? The India hiring cost calculator runs the same maths across any salary and role. Estimates for budgeting, not a quote. Assumes wages set at 50 percent of CTC in line with the labour codes, an employer electing to administer provident fund at the statutory wage ceiling, and a single employee with no dependants on the group policy. Professional tax is an employee-side deduction and is not included in employer cost. Currency converted at a fixed reference rate; ask us for a live figure.
04 — The part everyone gets wrong
Employer statutory cost falls as salary rises
This runs against intuition and against most of what is published about hiring in India, so it is worth being precise. Two of the three main employer obligations are capped in absolute terms rather than being a percentage of salary.
Provident fund is ceiling-linked
The statutory wage ceiling for provident fund purposes has stood at ₹15,000 a month since 2014. Employers may lawfully administer their contribution against that ceiling, which fixes the employer share at ₹1,800 a month regardless of whether the engineer earns ₹15 lakh or ₹80 lakh. Some employers voluntarily contribute on full wages. That is a choice, not an obligation, and it is worth making deliberately rather than by payroll default.
State insurance does not apply at all
Employees' State Insurance covers employees earning up to ₹21,000 gross a month, with the employer paying 3.25 percent. Every AI engineer you hire will be above that threshold, so the employer contribution is zero. Cost models that apply a flat "employer contributions" percentage to an AI salary are simply wrong.
Gratuity is the one that scales
Gratuity accrues at roughly 4.81 percent of wages and is uncapped, so it is the only significant employer obligation that grows with the salary. On a senior hire it is the overwhelming majority of your statutory cost. It is also the line the November 2025 labour codes moved.
The practical consequence
On a ₹40 lakh AI engineer, employer statutory contributions land near ₹1.2 lakh a year, close to 3 percent of CTC. Foreign employers routinely budget 15 to 20 percent on top of salary because that is the ratio in their home market. In India, at this salary level, that over-provisions the statutory line by a wide margin. The money is better spent on the salary itself, where it actually competes for the candidate.
05 — What changed
The November 2025 labour codes, and what they mean for a high-salary hire
India's four labour codes were notified to come into force from 21 November 2025, consolidating 29 earlier laws. The change that matters for AI hiring is the single definition of wages: basic pay, dearness allowance and retaining allowance must together be at least 50 percent of total remuneration, and where excluded allowances exceed that share, the excess is pulled back into wages for statutory purposes.
Indian salary structures were historically built the other way around, with basic pay held at 30 to 40 percent of CTC and the balance paid as allowances, precisely to shrink the statutory base. Those structures now have to be rebuilt. For a high-salary engineering hire the effect is narrow but real:
- Gratuity accrual rises sharply in proportion. Moving wages from 30 percent to 50 percent of a ₹40 lakh CTC lifts annual gratuity provisioning by roughly two thirds. Actuarial assessments of the change put gratuity liability increases at 25 to 50 percent or more for organisations that ran low-basic structures.
- Total employer cost moves much less. Because provident fund is ceiling-linked and state insurance does not apply, the same restructuring moves total employer cost by roughly one percent of CTC. Both facts are true at once, and most coverage reports only the first.
- Take-home falls unless you top up. A higher wages base means a larger employee provident fund deduction. If your offer was benchmarked on take-home, the candidate sees less than they expected. Model the net figure before you send the offer letter, not after.
- Fixed-term employees now accrue gratuity after one year rather than five, which changes the economics of short-term or project-based AI engagements considerably.
On the implementation date. The codes were notified with effect from 21 November 2025, but they operate through rules made at both central and state level, and states are at different stages of notifying theirs. Treating a single national switch-on date as settled is a mistake. Where you hire changes when parts of this bind, which is one more reason a five-city view beats a country-level one.
06 — The route
Entity, EOR or contractor
| Own entity | Employer of Record | Independent contractor | |
|---|---|---|---|
| Time to first hire | 3–6 months | Days | Days |
| Who is the legal employer | You | Saileor | Nobody, by design |
| Makes sense at | Roughly 25+ people, or where an India entity is strategic | 1–25 people, or testing the market | Genuinely project-scoped, short work |
| Misclassification exposure | None | None | High for full-time, directed work |
| IP assignment | Via employment contract | Assigned through to you by contract | Only if expressly assigned in writing |
If the EOR column is the one you are weighing, our Employer of Record in India page sets out exactly what the legal employer takes on and what stays with you. If you are not yet sure which of the three columns you belong in, the India Hiring Readiness Score asks eight questions and gives you a recommendation, a timeline and the compliance risks specific to your situation.
The IP point, which matters more for AI roles than for anything else
Under Indian copyright law, the default position for work made by an employee in the course of employment is that ownership sits with the employer. For an independent contractor the default runs the other way: absent an express written assignment, the creator can retain rights. For a support role that risk is abstract. For an engineer building your model architecture, training pipeline or retrieval system, it is the asset itself.
Two things follow. First, if you engage AI engineers as contractors, the written assignment clause is not boilerplate and should be reviewed by someone who has read Indian copyright law rather than copied from a US template. Second, if the engineer is working full time, to your direction, on your systems, calling them a contractor does not make them one. Indian authorities look at the substance of the relationship, and the exposure sits with you.
07 — Working with us
Why an India-only EOR for an India-only hire
Most EOR platforms cover 100 or more countries. That breadth is genuinely useful if you are hiring across eight of them. If you are hiring in India, it means your compliance questions route to a partner network and your answers arrive second-hand.
Saileor only does India. We run payroll on our own engine rather than reselling a platform, we hold ISO 27001:2022 and ISO 9001:2015, and pricing starts from $99 per employee per month as a flat fee that does not scale with salary. On a ₹40 lakh AI engineer, a percentage-of-salary fee is the single largest avoidable line in your cost stack. There is no lock-in and no setup cost.
Questions
Frequently asked
Is India really the second-largest AI talent pool in the world?
It depends entirely on what you are counting, and the claim is frequently overstated. India ranks second in the world for AI authors and inventors, behind the United States. On installed AI talent share it ranks third, at around 7 percent, behind the United States at roughly 33 percent and China at roughly 24 percent. On AI talent concentration as a share of the workforce it ranks around 13th, behind Israel, Singapore and South Korea. All three of these are real measurements of different things. Be sceptical of any page that quotes one without saying which.
Is Hyderabad the second-largest AI talent pool in the world?
No. Hyderabad is India's second-largest AI hiring market, behind Bengaluru, and that is where the claim comes from. On global city rankings of AI talent it has placed around 19th, with Bengaluru around 5th. Hyderabad's real advantages are different and more useful: it holds roughly 14 percent of India's digital tech talent and about 20 percent of the country's global capability centres, and it has been the fastest-growing GCC hub in the world. That produces a specific kind of engineer, trained inside large production organisations, at roughly 20 percent below Bengaluru pricing.
How long does it take to onboard an AI engineer in India through an EOR?
From signed offer to compliant employment, typically 48 hours once the candidate's documents are in. The constraint is almost never the employment mechanics. It is the notice period, which in Indian tech commonly runs 60 to 90 days at established employers. Factor that into your start date, not the onboarding time.
Do I have to contribute provident fund on the full salary?
No. The statutory obligation is calculated against a wage ceiling of ₹15,000 a month, which fixes the employer contribution at ₹1,800 a month. Contributing on full wages is permitted and some employers do it as a benefit, but it is not required. Decide it deliberately, because at AI salary levels the difference over a year is substantial.
Is the provident fund wage ceiling changing?
Possibly, and it is worth watching. The ceiling has stood at ₹15,000 since 2014. Press reporting indicates the finance ministry has approved raising it to ₹25,000, with a likely effective date well after the current financial year, subject to cabinet timelines. It was not changed in the most recent union budget. Until a notification is issued, ₹15,000 remains the operative figure. If the increase lands, employer cost on an AI engineer rises by roughly ₹12,000 a year, which does not change the shape of the decision.
Does professional tax make some Indian cities more expensive to hire in?
Not materially, and this is worth correcting because it gets repeated. Professional tax is a state levy deducted from the employee, constitutionally capped at ₹2,500 per person per year. Karnataka, Telangana, Maharashtra and Tamil Nadu levy it; Delhi, Haryana and Uttar Pradesh do not. So an NCR hire has one fewer registration and filing obligation than a Bengaluru hire, which is a compliance difference worth knowing about, but on a ₹40 lakh salary it is a rounding error and it is not your cost anyway. City cost differences in India are driven by salary, not by statute.
Can I hire an AI engineer in India as a contractor instead?
You can, if the engagement is genuinely project-scoped and the engineer controls how and when the work is done. If they work full time to your direction on your systems, the arrangement is employment in substance whatever the contract calls it, and the misclassification exposure sits with you. For AI roles there is a second reason to be careful: intellectual property created by a contractor is not automatically yours without an express written assignment.
When does it make sense to set up our own Indian entity instead?
Broadly, past 25 or so employees, or earlier if an Indian entity serves a purpose beyond employment, such as billing Indian customers, holding local IP or raising locally. Below that, incorporation typically takes three to six months and adds a permanent compliance function you have to staff. We will tell you when you have crossed the line. Losing a client to their own entity is a reasonable outcome; keeping one on an EOR they have outgrown is not.
Statutory figures on this page reflect the position as at the date shown below and are provided for general guidance. They are not legal or tax advice. India's labour codes are being operationalised through central and state rules on different timelines; confirm the position for your state before acting. Last reviewed: 11 August 2026