AI / Agentic Systems Engineer

About UnderAI

UnderAI is building AI-powered software for the insurance industry. Our platform works with large and complex insurance documents and helps brokers automate workflows such as policy audits, RFQ creation, risk analysis, document comparison, and policy placement.

We are looking for an engineer who has hands-on experience building *agentic AI systems in production*, not just calling an LLM API.

What You’ll Work On

You’ll work on production AI systems that reason over large documents, call tools, retrieve structured and unstructured information, and execute multi-step workflows reliably.

You will work on problems such as:

• Building and improving agentic workflows for insurance document processing.

• Designing agents that can use tools safely and reliably.

• Working with large-context documents such as insurance policies, placement slips, RFQs, and exposure documents.

• Building document extraction, retrieval, grounding, and comparison pipelines.

• Implementing human-in-the-loop workflows where users can review and approve AI decisions.

• Designing evals to measure agent accuracy, tool usage, regressions, hallucinations, and reliability.

• Improving prompt architecture, context management, and failure recovery.

• Building structured outputs and validation layers around LLM responses.

• Reducing hallucinations through grounding, deterministic checks, and verification systems.

• Working closely with product and engineering teams to move AI features from prototype to production.

What We’re Looking For

You should have experience with several of the following:

• Building LLM-powered or agentic systems in production.

• Tool calling / function calling.

• Agent orchestration and multi-step workflows.

• RAG, embeddings, vector search, or hybrid retrieval.

• Working with large documents or long-context LLMs.

• Structured output generation and schema validation.

• Prompt engineering and context engineering.

• LLM evaluation frameworks and regression testing.

• Guardrails and prompt-injection mitigation.

• Human-in-the-loop AI systems.

• Experience working with OpenAI, Anthropic, Gemini, or similar model APIs.

Engineering Skills

Strong experience with:

• TypeScript / Node.js or Python.

• REST APIs and backend development.

• PostgreSQL or similar relational databases.

• Git and modern software engineering workflows.

Experience with the following is a plus:

• Next.js

• Drizzle / Prisma

• LangGraph / LangChain or similar orchestration frameworks

• Vector databases

• PDF/document parsing

• OCR and document intelligence

• Background job systems and queues

• Docker

• Cloud deployments

What Makes You Stand Out

You’ve built an AI system where:

• The model had access to tools.

• Reliability mattered more than just producing a good demo.

• You had to handle hallucinations and failures.

• You built evals to measure performance.

• You thought about prompt injection or tool security.

• You had to reason over large amounts of context.

We care significantly more about *what you have built* than certifications or buzzwords.

Please include GitHub repositories, demos, or projects showing agentic systems you have worked on.

Pay: ₹239,078.31 - ₹807,927.24 per year

Work Location: In person