# AI & Full-Stack Developer — 2026 Graduates Location: T-Hub, Gachibowli, Hyderabad (Work-from-Office) Duration: 3 Months Stipend: ₹15,000/month Schedule: Monday to Saturday | 10:00 AM – 6:00 PM Company: Altibbe Health Pvt. Ltd.

Product: HEDAMO Apply to: [email protected] Subject Line: Application — Graduate Full-Stack & AI Developer | 2026 Eligibility: 2026 pass-out only — degree completed and available to join immediately. --- ## About the Role Altibbe is hiring a 2026 graduate for a hands-on technical role at T-Hub, Gachibowli, Hyderabad. We’re looking for someone who can use modern AI engineering tools to build, automate, test, document, deploy, and improve real software systems. This is not a conventional internship built around observation or training exercises.

The selected candidate will work on live products and internal systems from day one. You’ll work on:

- AI agents and agentic workflows
- Workflow and task automation
- Full-stack product development
- Website testing and continuous improvement
- APIs, databases, and integrations
- Cloud infrastructure and deployment
- Terminal-based engineering and debugging
- Technical documentation and knowledge repositories --- ## Tech Stack & Engineering Environment Frontend: React, Next.js, TypeScript/JavaScript, Tailwind CSS Backend & Data: Node.js, REST APIs, PostgreSQL, database design and migrations, authentication AI & Automation: OpenAI, Anthropic or equivalent LLM APIs, prompt and context engineering, tool/function calling, structured outputs, agentic workflows, automation pipelines, retrieval and document-processing workflows Cloud & Engineering Operations: GCP or equivalent, Linux, shell/SSH, Git/GitHub, Docker, environment variables and secrets, logs and production debugging, CI/CD fundamentals AI Coding Tools: Claude, Claude Code, ChatGPT, Codex, Cursor, GitHub Copilot, or equivalent --- ## Who We're Looking For You don’t need years of professional experience. You need evidence that you can build.

You should be able to:

- Understand an unfamiliar codebase.
- Use AI coding agents without blindly accepting their output.
- Break a problem into steps and execute it.
- Build an automation or AI workflow from scratch.
- Work confidently in a terminal.
- Connect APIs and services.
- Test your own work.
- Diagnose failures using logs and evidence.
- Write clear technical documentation.
- Identify improvements in an existing product and implement them.
- Communicate what you did, what failed, and what remains uncertain. Personal projects, GitHub work, hackathons, deployed applications, AI experiments, automations, open-source contributions, and independently built tools are strongly valued. --- ## Eligibility You must be:
- A 2026 pass-out only
- Degree completed
- Available to join immediately
- Able to work from T-Hub, Gachibowli, Hyderabad
- B.Tech / BE / BSc / BCA / MCA / M.Tech or equivalent --- # Selection Process The selection process begins with a practical AI engineering assignment. Stage 1: Practical AI Engineering Assignment Stage 2: Short Validation Call Stage 3: In-Person Technical Assessment Day at T-Hub for shortlisted candidates --- # Stage 1 — Practical AI Engineering Assignment ## Build an Agentic Work Intake & Execution Prototype ### Objective Build a small but working AI application that turns unstructured incoming work — such as an email, meeting notes, founder instruction, customer request, or bug report — into a structured, reviewable, partially automated workflow.



This is a build exercise, not a design exercise. Submit working software, not a document. --- ## Your Prototype Must Include ### 1.

Intake

Provide a simple way to submit unstructured text. A web interface is preferred, but a clear CLI is acceptable. ### 2. AI Understanding Use an LLM to extract information using a defined structured schema, not free-form prose.

Extract

- Task title
- Summary
- Action items
- Priority
- Detected deadline
- Missing information
- What could be automated
- What requires human confirmation ### 3.

Agentic Planning

Generate an execution plan that routes each action as:

- Execute automatically
- Prepare for human review
- Cannot execute with available tools
- Requires clarification Include a brief reason for each decision. ### 4. Real Tools / Functions Implement at least three real tools or functions.

Examples include

- Draft a communication
- Create a task record in persistent storage
- Generate a Markdown brief
- Run a bounded website check
- Simulate a reminder — no real calendar invites
- Search stored work ### 5. Human-in-the-Loop Control At least one action must include an explicit: Approve / Reject / Edit step before the action is treated as complete. Do not send real external email. ### 6.

Persistence

Retain state between runs. SQLite is acceptable.

Store

- Original request
- Structured interpretation
- Action items
- Status
- Outputs
- Timestamps ### 7.

Activity Trace

Provide a visible log showing what the system did. The activity trace should be understandable without reading the source code. ### 8.

Failure Handling

Demonstrate at least one sensible failure path. The system should fail clearly rather than pretending to succeed. --- # AI & Agent Requirements Use OpenAI, Anthropic, Gemini, or another accessible LLM.

Your implementation should demonstrate

- Structured output / JSON schema
- Tool or function calling
- Multi-step execution with state passed between steps
- Validation
- Explicit agent boundaries Frameworks such as LangChain, LangGraph, CrewAI, or similar are permitted but not required. A simple implementation you understand is better than a complex implementation you cannot explain. --- # AI Coding Tools AI coding tools are encouraged and will not be penalized. Include a short "How I Used AI" section in your README covering:
- Tools used
- What you used them for
- One example of an AI mistake
- How you identified and fixed that mistake --- # Terminal & Cloud Requirements Provide the exact terminal commands required to:
- Clone
- Install
- Configure
- Initialize
- Run
- Test your submission. A live deployment is strongly preferred. Never commit API keys or credentials. --- # README Requirements Your `README.md` must explain:
- What the application does
- Architecture — a Mermaid diagram is welcome
- Agent workflow: Intake → Interpretation → Planning → Tools → Approval → Persistence → Completion - Setup instructions
- Environment variables — placeholders only
- Design decisions
- Limitations
- What you would build next — maximum 5 items




- How you used AI --- # Required Test Scenarios Test your application against these three fixed scenarios and include evidence of the outputs. ## Scenario 1 — Routine Business Work Summarize a partner discussion, extract follow-ups, draft a thank-you email, and set a 7-day reminder. ## Scenario 2 — Product / Website Work Review hedamo.com, run whatever automated checks your prototype actually supports, and produce a short technical report. Do not claim checks that your system cannot actually perform. ## Scenario 3 — Ambiguous Request Use this request: > "Please take care of the documentation and send it to everyone before the meeting." The system should identify and flag missing information rather than inventing recipients, documents, or meeting details. --- # What to Submit Your submission must include:
- GitHub repository URL
- Live prototype URL, if deployed
- README.md
- Sample outputs from all three required test scenarios
- A screen recording of 5 minutes or less
- Resume in PDF
- GitHub profile / portfolio, if available Your screen recording should demonstrate:
- The application
- One complete workflow
- At least one tool call
- The approval step
- The activity trace
- The project running from your terminal --- # Time Expectation Approximately 3–5 hours. A smaller prototype where every part genuinely works is better than a larger project with broken features. AI coding tools and open-source libraries are welcome. The work must still be something you understand and can explain, not an unchanged tutorial project. --- # How We'll Evaluate It Agentic & Automation Thinking: 25% Technical Execution: 20% AI Engineering Judgment: 15% Automation Quality: 15% Terminal / Cloud / Engineering Operations: 10% Documentation: 10% Product Judgment & Attention to Detail: 5% ### Notice what's missing: memorisation. We care about your ability to build, reason, automate, test, debug, and explain. --- # A Submission May Be Rejected If - It is a document or mockup rather than working software.
- The repository does not run using your provided instructions.
- Secrets or API keys are committed publicly.
- You cannot explain your own code.
- Outputs are fabricated or manually produced but presented as automated.
- The application silently claims success after a failure.
- The required approval step is skipped.
- There is no meaningful AI or agentic component.
- The project is substantially an unchanged tutorial.
- You are not a 2026 pass-out available to join immediately. --- # Code + Judgment We’re looking for people who can combine software engineering, AI tools, automation, and sound technical judgment.

Solid candidates

- Build rather than only discuss.
- Use AI as an engineering multiplier, not a substitute for understanding.
- Verify AI-generated code and outputs.
- Test their own assumptions.
- Diagnose problems using evidence and logs.
- Understand where automation should stop and human judgment should begin.
- Communicate limitations and uncertainty clearly.
- Care about reliability, security, and maintainability. --- # Ready to Build? This role is for a 2026 graduate who wants to work directly on AI agents, automation, full-stack systems, cloud infrastructure, and production software. Show us what you can build. ## How to Apply Email your application to: [email protected] Subject Line: Application — Graduate Full-Stack & AI Developer | 2026 Include:
- Resume
- Assignment repository
- Prototype URL, if deployed
- 5-minute demo video
- GitHub / portfolio, if available
- A 3–5 line introduction

📌 AI & Full Stack Intern (Hyderabad)
🏢 Altibbe Health Private
📍 Hyderabad