Remote · 3rd & 4th year B.Tech CS
AI Engineering
Internship
Work on agentic AI systems, end to end. Your code goes into products customers actually use — in the US, UK and India.
Only 4 places this cycle. Starting immediately.
Earn up to
₹30,000
or $300 / £300
a monththree work packets, going flat out
₹10,000 / $100 / £100 per packet — about 10 days of focused work. Stretch to 45 days if you're busy; the pace is yours.
Done means merged, tested, reviewed and running — then you're paid.
Internship certificate — after you finish three packets. We will write you a reference too.
How the work packet model works
We don't pay by the hour and we won't ask you to clock in. We scope a discrete piece of real work — a work packet — write down what "done" means before you start, and pay ₹10,000 / $100 / £100 when you deliver it.
A packet is about 10 days of focused work — our ballpark, not a target. Give it your full attention and you'll finish in roughly that; work efficiently and it takes less. Mid-semester with exams stacking up, take up to 45 days. The amount of work never changes — only your pace does, and you can take as many packets as you can.
This only works because the definition of done is unambiguous, so it always is: merged, tested, reviewed and running. You never have to guess whether you're finished, and we never have to look over your shoulder to find out.
A packet is one piece of a bigger system, and you need to understand the whole system to build your piece properly. So before you start, we give you the project documentation and walk you through what the product does, what problem it solves, and where your packet fits.
This matters more than it sounds. Code that works on its own but misses the point of the project is the most common way this goes wrong. If the bigger picture is not clear to you, ask — explaining it is our job, not something for you to guess.
What you'll build
- LLM orchestration with LangGraphMulti-step agents, tool calling, and the state machines that keep them predictable.
- RAG and vector retrievalChunking, embeddings, and measuring whether retrieval is actually returning the right thing.
- EvalsDeciding whether a change made the system better, with numbers rather than vibes.
- Real-time voiceStreaming speech in and out, inside a latency budget a caller will tolerate.
- AWS infrastructureLambda, ECS and the deployment path that puts your work in front of customers.
Worth being straight about: most of this is ordinary software engineering. Clean Python, sensible structure, tests, code review, deployments. The AI sits on top of a normal, well-built codebase — and the codebase is the part that has to be right.
Must have
- Strong PythonReal software engineering, not just notebooks.
- You understand the code you writeAI assistance is fine and we use it too. Shipping code you cannot explain is not.
- An AI project you've builtA RAG app, an agent, an LLM tool — something you can walk us through.
- Cloud comfortAWS ideally; any cloud counts. A deployed side project is plenty.
Bonus — none of these are required
How to apply
- 01Send us your CVWhatsApp or email, with links to anything you've built.
- 02A short callThirty minutes. Mostly you walking us through one thing you made.
- 03We scope your first packet togetherYou see the work, the definition of done, and the fee before you commit to anything.
- 04Build it, ship it, get paid₹10,000 / $100 / £100 per packet on delivery. Then pick up the next one.
Questions
What exactly is a work packet?
A standard, quantifiable piece of real work — a feature, an integration, a service — with an agreed definition of done written down before you start. We size a packet at roughly ten days end to end: built, tested, and working properly, not merely written. That is our ballpark rather than a target — if you are efficient, it takes less. You know what finished looks like on day one, which is what makes it possible to work without someone supervising you daily.
How much can I actually earn?
₹10,000 / $100 / £100 per work packet, and there is no cap on how many you take — as many as you can. We size a packet at roughly ten days of focused work, so a student who is genuinely free, over a break say, might finish three in a month and earn ₹30,000 / $300 / £300; work faster and it could be more. During a normal semester, with classes and exams, one a month is a more realistic pace. You choose.
How many hours a week does this take?
That is up to you. We size a packet at about ten days of focused work — our estimate, and often less once you know the codebase. Whether you compress that into a week or spread it across six is your call. Most students in semester find 15–20 hours a week comfortable.
Can I do this alongside classes and exams?
Yes — that is the entire reason the model exists. When exams hit you can stretch a packet to 45 days without anyone chasing you. The quantity of work never changes; only your pace does.
What if I can't finish a packet in 45 days?
Tell us early. We would much rather re-scope a packet than have someone go quiet. Deadlines slip for real reasons and we have all had a semester go sideways. What does not work is disappearing.
Do I need prior AI experience?
You need to have built something with LLM APIs — a RAG app, an agent, a tool — and be able to explain how it works and what you would do differently. Coursework counts. A side project counts. A tutorial you followed without understanding does not.
What if I use AI to write the code?
So do we. What matters is that you understand every line you ship and can defend it in review. If you cannot explain why the code is the way it is, it is not ready — and that will show up quickly in a review.
Is it fully remote?
Yes. Fully remote, no relocation, no fixed hours. We work with clients in the US, UK and India, so some flexibility on call times helps, but there is no expectation you are online at 9am.
How many people are you taking?
Four in this cycle, starting immediately. We keep it small deliberately — every packet gets reviewed properly, and that does not scale to twenty interns.
Will I get an internship certificate?
Yes — on successful completion of at least three work packets, meaning three packets that met the agreed definition of done. It comes with a reference we will actually stand behind, because we will have read your code.
Could this turn into a full-time role?
It can, and we would like it to for the right people. It is not promised. The honest version: do good work across a few packets and a conversation becomes very likely.
How and when do I get paid?
Per packet, on delivery — once the work is merged, tested, reviewed and running. Bank transfer to your account. No waiting until the end of a three-month term.
To apply
Send us your CV and links to anything you've built.
AIQuest builds production AI for businesses in the US, UK and India. Founded by Georgia Tech alumni.