Agentic AI Engineer

Agentic AI Engineer (Flutter / Next.js)

Location: Hyderabad | Type: Full-time

About the role

Were looking for a self-driven Agentic AI Engineer who ships real products fast using Claude and strong prompt engineering. You’ll learn on your own, research what’s needed, and build, publish and deploy mobile apps, web apps and AI agents end to end: clean UI, backend, integrations and deployment, without waiting for someone to tell you what to do.

Must-have

• Claude expertise (mandatory): you use Claude daily and know it well: Claude chat, Claude Code, the Claude API, and how to pick the right approach for each task. You know how it behaves, where it fails and how to get it back on track.
• Prompt engineering (mandatory): you write clear, structured prompts with context, examples and constraints. You break big tasks into steps, ask for a plan before code, and fix bad output. You can show us real prompts you’ve used.
• Good UI skills: you build clean, responsive, user-friendly screens. You care about layout, spacing, usability and how the product feels on mobile and web. Show us UIs you’ve built.
• Backend skills: you build APIs and server logic (Node.js or similar) with MongoDB. You handle authentication, data models, error handling and security basics.
• Integration skills: you connect third-party services and APIs: payment gateways, WhatsApp, maps, notifications, webhooks and LLM APIs. You can read API docs and get an integration working on your own.
• Self-learner and independent: when you hit something new (a tool, API, framework, error), you research it, learn it and solve it yourself. You take a task from idea to published, end to end.
• Shipped work: at least 1 app and 1 website deployed, built using AI tools
• Languages: JavaScript and Dart
• Frameworks: hands-on experience in Flutter, Next.js, or both
• Mobile: at least 1 app published to the Play Store or App Store. If not published, at least 1 app fully built (share the repo or APK/TestFlight)
• Agentic AI: experience building and publishing AI agents with LLM APIs, tool/function calling or MCP
• Communication (important): clear spoken and written English. You explain what you built, why, and what didn’t work, in plain words. You are honest about what you know and don’t know. You say “I’m not sure, let me check” instead of guessing, you take feedback well, and you work humbly with a small team. We are not looking for overconfident people who claim to know everything.

Good to have

• REST API design and Node.js
• CI/CD (Codemagic, GitHub Actions, Vercel)
• WhatsApp or chatbot integrations
• MCP servers, Claude Skills or Claude Projects

What you’ll do

• Build and ship Flutter apps and Next.js web apps with good UI
• Build backend APIs and connect them to MongoDB
• Integrate payments, WhatsApp, maps and other third-party services
• Research, build and publish AI agents on your own, from first idea to live
• Use Claude daily to move faster, and help the team prompt better
• Deploy to the Play Store, App Store and web
• Own features end to end, with no hand-holding
• Tell the team early when something is blocked or going wrong

How to apply

Send:

1. Links to your deployed website and published app (or repo/APK if unpublished)
2. A 2-minute video or short note on how you used Claude to build them
3. One real prompt example: the prompt you wrote, what Claude returned, and how you improved it
4. One integration you built: which service, what was hard, how you solved it
5. One thing you taught yourself: a tool or topic you didn’t know, how you learned it and what you shipped with it
6. One mistake: something that went wrong in a project and what you learned from it
7. Your GitHub / portfolio

Shortlisting (candidates without Claude and prompting skills will not pass):

1. Short call to check communication and attitude
2. Live Claude task (30-45 min): we give you a feature to build end to end that they haven’t built before, for example a screen with good UI, a small backend API on MongoDB, and one integration. Use Claude and any resources you like. We watch how you prompt, how you handle bad output, how you review code before accepting it, and whether you get it working without asking us for help.

Screening questions for the call:

• “How do you start a prompt for a big feature?” (Strong: context, constraints, plan first, small steps)
• “Claude gave you wrong code. What do you do?” (Strong: share the error, narrow the scope, test it)
• “When do you use Claude Code vs chat vs the API?”
• “Walk me through an integration you built. What broke?” (Strong: specific details, real problems)
• “How do you design a MongoDB collection for an orders feature?” (Strong: thinks about fields, indexes, status flow)
• “Tell me about something you got wrong.” (Strong: owns it, explains the fix. Weak: blames the tool or others)
• “What don’t you know well yet?” (Strong: names real gaps. Weak: “nothing, I can do everything”)

Spotting overconfidence: they claim to know everything, never say “I don’t know”, can’t name a mistake, or dismiss feedback. Good candidates are confident in what they’ve shipped but honest about their limits