Engineer - Agentic AI

TIH | IIT Bombay; Mumbai; On-site; INR 1,800,000 - 3,400,000; Full time

Job summary

TIH | IIT Bombay invites experienced software/AI engineers to join a team building agentic AI systems that plan, act, and self-correct to automate complex enterprise processes.

You will own end-to-end delivery—from frontend UIs to backend services and deployment—shaping scalable, production-grade AI workflows with strong emphasis on reliability and safety.

Qualifications

• 3–5 years of professional software/AI engineering experience building agentic systems.
• Hands-on experience with multi-step planning and tool usage by agents.
• Experience with RAG pipelines and grounding techniques.
• Familiarity with local/self-hosted LLMs and efficient inference.
• Experience in end-to-end SDLC and production-grade AI systems.

Responsibilities

• Design and build agentic AI workflows and systems that plan, act, and self-correct.
• Identify and re-imagine repetitive or complex workflows as agentic solutions.
• Develop agent orchestration and integrations with enterprise systems.
• Deliver complete end-to-end solutions: UI, backend, and deployment.
• Implement RAG pipelines to ground agents in enterprise data.
• Deploy and operate self-hosted LLMs with scalable inference.
• Ensure AI systems are scalable, observable, and safe with guardrails.

Skills

Agentic AI engineering

Orchestration frameworks

MCP protocol

Full-stack delivery

Prompt engineering

Tools

LangChain

LangGraph

LlamaIndex

AutoGen

CrewAI

Job description

• Design and build agentic AI workflows & systems — multi-step, tool-using agents that plan, act, and self-correct to automate enterprise processes.
• Identify manual, repetitive, or complex business workflows and re-imagine them as reliable agentic solutions.
• Build and maintain agent orchestration and the tools, connectors, and integrations that let agents act on real enterprise systems.
• Build and own complete solutions end-to-end — front end, backend, and deployment — not just the AI/agent layer. Expect full-stack ownership.
• Design and implement RAG pipelines to ground agents in enterprise knowledge and data.
• Deploy and operate local / self-hosted LLMs, with a focus on fast, efficient inference at scale.
• Build and expose capabilities via MCP (Model Context Protocol) for tool and context integration.
• Engineer human-in-the-loop review and approval flows for high-stakes automated actions.
• Ensure AI systems are scalable, reliable, observable, and safe in production — with guardrails, validation, retries, graceful failure handling, and cost/budget caps to prevent runaway spend.
• Follow solid SDLC practices end to end: version control, code review, CI/CD, testing, and deployment.

Minimum Qualifications and Experience:

• 3–5 years of professional software / AI engineering experience, with demonstrable hands-on work building Agentic AI systems — not just calling an LLM API but building agents that plan and act over multiple steps.

Required Skills:

• Hands-on experience with agent frameworks and orchestration — e.g., LangGraph, LangChain, LlamaIndex, AutoGen, CrewAI, or similar.
• Solid understanding of agentic harnesses / scaffolding — the execution loop, tool interfaces, context and memory management, and evaluation that make an LLM behave as a reliable agent.
• Proven experience building RAG systems — chunking, embeddings, vector databases, retrieval strategies, and grounding.
• Basic familiarity with local / self-hosted LLMs — faster inference, quantization techniques, and serving frameworks such as vLLM (a plus, not core).
• Hands-on experience with MCP (Model Context Protocol) for tool / context integration.
• Full-stack delivery experience — comfortable building a usable UI (React/Next.js or Streamlit) and a backend/API layer (FastAPI or Node.js) end-to-end.
• Strong grasp of core LLM application patterns: prompt engineering, tool / function calling, structured (JSON) output handling, context management, and evaluation.
• Comfort debugging non-deterministic systems and designing guardrails, validation, and fallbacks.
• Proficiency across traditional ML, NLP, Computer Vision, and Deep Learning fundamentals & techniques.
• A clear track record of deploying scalable AI solutions in production, not just building POCs.