AI builder internship · Paid · Remote, India

Show us something you built. Then build with us.

An internship on Zuddl's internal tools, not the product customers see. You'll build the AI tools and automation that Zuddl's own teams run on. Two steps, no aptitude test, no puzzle round: send us a project you already built , and if you're shortlisted, spend a few days building with us on a real Zuddl problem.

Submit your project What the role is

Step 1 · Your application Three things
Your resume PDF. One page is plenty.
A public GitHub repo Built recently, using AI. The README should say how to run it.
A 3–6 minute demo video You showing it working and how it's built. A screen recording with your voice is fine. Not a slide deck.
Then a short form about the project and how you work. Allow about half an hour, plus the video. You'll hear back either way. If you're shortlisted, step 2 is a few days building with us.

Role: AI builder intern, internal tooling
Duration: 3 months, paid, extendable to 6
Location: Remote, anywhere in India
Who: 2nd year to recent grad
Intake: Applications always open

You build tools for the people you work with.

Zuddl is an AI-native events platform. This role sits one layer behind it: you work with Zuddl's own engineering, design, marketing and revenue teams on the internal tools and automation they use every day.

This isn't for you if

You want to work on the product customers see.
You'd rather stay quiet until you're sure you're right.

About the second one. Say your ideas out loud, even the half-formed ones. Staying quiet until you're sure doesn't work here.

A short list, and an honest one.

You'll have a coding agent for the rest. What we can't give you is the judgement to know whether what it produced is right.

Non-negotiable

Databases Model data for a problem, query it, and say why you shaped it that way.
APIs Consume one and build one.
Debugging Work out why something broke, instead of re-prompting until the error goes away.

What we're looking for

A CS background, ideally Not required. Understanding a working stack counts for more than where you learned it.
Opinions about your tools Claude Code, Cursor, Codex, whichever. We care less which you picked than whether you can say why, and where it's wrong.
Recent work The tooling moved a long way in the last few months. Show us work done on the current version of it.
Things you run yourself Agents, workflows or assistants you've wired up beyond a chat window. Hands-on only; we'll ask about whatever you mention.

We don't ask for

Aptitude tests or puzzle rounds We don't run them, at any stage.
A formal degree If the understanding is real, we don't care where it came from.
Anything about events Your project doesn't need to touch events or Zuddl's API. Any stack, any domain.

Someone should be thankful it exists.

That's the whole test, and the someone can be you. Some common portfolio projects are an automatic no, and we'd rather say so here than in a rejection email.

Automatic no

A demo you followed from a YouTube tutorial. It tells us nothing about you.
A non-functional clone of an existing site. If it doesn't work, it isn't a product.
A to-do list , or anything of that kind.

What we want to see

Scripts that automate real work. Something that was done by hand, and now isn't.
A tool that serves even one person. A friend, a teacher, a shop, or you, as long as it's actually used. A command-line tool counts.
Anything someone is thankful for , or would be if they knew it existed.

Small is fine. So is ugly, and so is an audience of one.

Four things we read for.

Not polish, and not how many models you chained together.

A real problem, and the people who had it: Something manual, slow or broken for specific people. Five users counts. "Nobody yet" is allowed; it just makes the other three matter more.
Judgement about where AI belongs: The model does what only a model could, the rest is deterministic on purpose, and you can say what it gets wrong.
Taste in how you built it: Which tools, for what, and one thing they got wrong that you caught.
Fundamentals under the hood: A data model that fits the domain, code someone else could read, a README that runs, no secrets in the repo.

Using AI to build it is expected, not a confession. We want to know where you disagreed with it.

Step 2 · If you're shortlisted

A few days on a problem we actually have.

Shortlisted candidates spend a few days on a real, unsolved Zuddl problem, working with the person who has it. It ends with a demo and a conversation about how you built it. Details come with the invitation.

Things people ask.

What counts as "built with AI"?

Either or both: the product uses a model at runtime, or you used AI tools to build it. Tell us which. Using Cursor, Claude Code, Copilot or similar is expected, not a mark against you.

Can I submit a team or college project?

Yes, if you tell us clearly which parts were yours. We'll ask about those parts.

What about the internship itself?

Three months, paid, remote within India, extendable to six by mutual agreement. Hours are flexible. You get a named mentor and a project of your own from week two. The stipend is competitive and is confirmed in the offer.

Your code speaks first.

If you've built something that made a real process better for real people, we want to see it.

Submit your project