Role & responsibilities

We are hiring a small number of early-career engineers who already build software with AI agents as part of how they design, write, test and ship code. You will take business problems we have identified with clients, build the harness that solves them (agents, tools, evaluations, integrations and the application around them), and put it in front of the people who own the problem. That often means working directly with CXOs, explaining what you built and changing it based on what they tell you.

The role is junior only in years. We expect your work to hold up next to engineers with far more experience.

What you will do

- Turn a loosely stated business problem into a working system: scope it, build a first version within days, and iterate with the client.
- Build agent harnesses: tool definitions, context and retrieval, orchestration, guardrails, evaluation suites, and logging that shows what the system actually did.
- Use AI coding agents as your default way of working, and know where their output needs checking.
- Write the APIs, data pipelines and front ends needed to make an AI component usable end to end.
- Choose models and approaches based on measured accuracy, latency and cost for the task.
- Present work to senior client stakeholders and turn their feedback into changes.
- Write tests and evals that show a system works before it reaches a client.
- Contribute to how the team practises AI-native development. Much of it is still being worked out.

What we look for

Required

- Less than 4 years of professional experience, internships included. Exceptional graduates are welcome.
- Strong programming in Python, TypeScript, or both.
- At least one LLM-based system you have built and shipped beyond a tutorial: an agent, a RAG pipeline, an automation, a tool that real people use. Show us.
- Daily use of AI coding tools (Claude Code,



Cursor, Codex or similar) and a view on how to use them well.
- Working knowledge of APIs, relational and NoSQL databases, Git, and deployment on AWS, Azure or GCP.
- Transparent written and spoken English. You can explain a technical decision to a non-technical executive.
- You work independently with minimal guidance. Given a brief, you find what you need, make sensible decisions on your own, and raise only the questions that need someone else's answer.

Valued

- Experience with MCP, tool calling, structured outputs, evaluation frameworks or multi-agent orchestration.
- Front-end work in React or similar.
- Document processing: OCR and extraction from PDFs, forms, financial or insurance documents.
- Open-source contributions, hackathon results, competitive programming, or side projects with users.
- A degree in computer science or a related field. Strong evidence of ability counts for more.

Compensation There is no fixed band. Compensation is set on what you demonstrate during the process: the quality of what you have built, how you perform on the practical exercise, and how you handle ambiguity and changing requirements. Candidates who perform strongly will be paid well above the market rate for their experience level.

Before you apply

This role is demanding. Problems arrive half-defined, priorities move with client needs, and you are accountable for outcomes. There is little handholding: you are expected to take a problem and run with it. Apply only if you are up to that.

The process:

1. Apply with a link to work you have built (GitHub, a live demo, or a write-up). Applications without one will not be reviewed.
2. A technical conversation about that work.
3. A build exercise on a realistic business problem, take-home or live. AI tools are allowed and expected.
4. A final conversation with leadership.

Location: Chennai, on-site

📌 Junior Software Engineer, AI-Native Development (Chennai)
🏢 Perfecter.ai
📍 Chennai