Hiring an AI engineer in India: time zones, rates and how to vet one
What it's really like to hire an AI engineer from India: how the time zones work, what rates look like, and the questions that separate builders from tinkerers.
I'm an AI engineer in Surat, India, and most of my clients are in the US, the UK and Europe. So I get asked the same few questions on almost every first call. Will the time zone be a problem? What should this cost? How do I know you're any good?
Here are honest answers, including the parts that don't flatter freelancers like me.
How do the time zones actually work?
India runs on one time zone, IST, which is UTC+5:30. Here's what a normal 9 to 6 day in a few client cities looks like on my clock:
| Question | Their 9am to 6pm, in IST | Easy overlap |
|---|---|---|
| San Francisco | 9:30pm to 6:30am | Their morning, my late evening |
| New York | 6:30pm to 3:30am | Their morning, my evening |
| London | 1:30pm to 10:30pm | Most of the day |
| Berlin | 12:30pm to 9:30pm | Most of the day |
| Dubai | 10:30am to 7:30pm | Practically the same day |
| Sydney | 4:30am to 1:30pm | My morning, their afternoon |
For European teams it barely matters. For US teams, the pattern that works is a short call in your morning, then a full day of work happening while you're offline. You wake up to a pull request, not a question.
What doesn't work is expecting someone in India to keep your exact hours every day for months. People who promise that burn out, and their work shows it. Ask for a fixed overlap window instead, and protect it.
What should it cost?
Rates in India range more than anywhere I know. You'll find people offering $10 an hour and people charging what a US contractor charges.
From what I see on the platforms, senior freelancers with real production AI work usually sit somewhere around $40 to $100 an hour, or quote a fixed price per project. Agencies cost more, because you're paying for a sales team and a project manager as well.
The cheap end isn't always a bad deal. But a lower rate often means one of three things: less experience, someone juggling several clients at once, or you doing the quality control yourself. Count your own review hours in the price.
How do you vet an AI engineer?
AI has made this harder. Anyone can build an impressive demo in an afternoon now. What's rare is someone who has kept a system running after real users found its weak spots.
These five questions sort people out quickly:
- "Walk me through something you put into production." Listen for real users, real data and real problems. A notebook demo is not an answer.
- "How did you know it was working?" You want to hear about test sets, logs, metrics. "The client was happy" is nice, but it isn't evidence.
- "What broke after launch?" Everyone has an answer to this. People without one haven't shipped.
- "How would you keep costs under control?" A good engineer talks about model choice, caching and measuring before optimising.
- "What wouldn't you use AI for here?" The best ones will talk you out of part of your own idea.
What does a good hiring process look like?
- 0130-minute callYour problem, their questions
- 02Short written planScope, price, what's out
- 03Small paid taskOne to two weeks
- 04First milestoneReal users, real data
- 05OngoingRetainer or next phase
The written plan matters more than it looks. If someone can't put the scope on one page after a call, the project will drift.
What about contracts, payment and code ownership?
- Platform or direct. Upwork and similar platforms handle payment, escrow and disputes, which is reassuring for a first project. Many clients move to a direct contract later, once there's trust on both sides.
- Currency. Most Indian freelancers invoice in USD, GBP or EUR, so there's no currency risk on your side.
- Code ownership. Make sure the contract assigns the IP to you, and that the code lives in your repository and your cloud account from day one. Not theirs.
- Data. Agree up front where your data may go, especially if it's sent to a model provider.
Red flags worth walking away from
- They agree to everything on the first call, including the deadline.
- Every past project is "under NDA" and they can't describe a single one, even in outline.
- They want to start building before anyone has written down what "done" means.
- The quote is a single number with no breakdown.
If you're weighing up an AI project and want an honest opinion on scope, even if you end up hiring someone else, send me a note.