AI Engineer (RAG & Agents)

Sabrixa; Gurugram District; On-site; INR 2,009,000 - 2,678,000; Full time

Job summary

Sabrixa is seeking engineers to build grounded RAG systems and agent workflows that stay accurate and avoid hallucinations. You will design retrieval pipelines, integrate LLMs with documents, and collaborate with engineering to ship real AI products with explicit human approval steps.

The role emphasizes hands-on API work, TypeScript or Python for service code, and documenting failure modes for reliable production systems.

Qualifications

• Hands-on experience with LLM APIs and at least one RAG stack.
• Proficiency in Python and/or TypeScript for service code.
• Understanding of chunking, embeddings, evals, and prompt failure modes.
• Bias toward grounding answers in customer-approved sources.

Responsibilities

• Design retrieval pipelines, evaluation sets, and tool-using agents for client and product use cases.
• Integrate LLMs safely with documents, structured data, and human approval steps.
• Partner with engineering to ship AI into real products (not demos only).
• Document failure modes, costs, and monitoring for production systems.

Skills

LLM APIs

RAG stack

Python

TypeScript

Chunking

Embeddings

Evals

Prompt failure modes

Grounding sources

Tools

LangGraph

AWS

Azure

GCP

WhatsApp

Document automation

Job description

Build grounded RAG systems, agent workflows, and AI features that stay accurate, confirm-before-save, not hallucinate-and-hope.

Key Responsibilities:

• Design retrieval pipelines, evaluation sets, and tool-using agents for client and product use cases
• Integrate LLMs safely with documents, structured data, and human approval steps
• Partner with engineering to ship AI into real products (not demos only)
• Document failure modes, costs, and monitoring for production systems

Requirements & Standards:

• Hands-on experience with LLM APIs and at least one RAG stack
• Python and/or TypeScript for service code
• Understanding of chunking, embeddings, evals, and prompt failure modes
• Bias toward grounding answers in customer-approved sources

Preferred Multipliers:

• LangGraph / agent orchestration experience
• AWS, Azure, or GCP deployment for AI services
• Prior work on WhatsApp or document automation

Sabrixa AI-Era Interview Protocol:

Zero whiteboard algorithm trivia. We evaluate proof of work (GitHub repos and live deployments), conduct a 45-minute architectural dialogue, run a 100% paid collaborative pairing micro-sprint, and deliver a founder decision within 48 hours.