Senior AI Engineer — Agentic Systems

Glorious Insight; India; On-site; INR 4,000,000 - 7,000,000; Full time

Job summary

Glorious Insight is seeking an experienced software engineer to design and ship production-grade AI systems. You will build multi-agent orchestration with guardrails and human-in-the-loop controls, and design retrieval pipelines for enterprise clients.

You will own evaluation frameworks, manage golden datasets, and integrate CI regression suites. You will work directly with client engineering teams from design to production and contribute to platform growth.

Qualifications

• 4+ years software engineering experience.
• 2+ years shipping LLM/ML systems to production.
• Strong Python programming skills and system design
• Experience with vector stores, inference serving, and cloud platforms (Azure/AWS).
• Bias for measurement and verifiable evaluation in deployments.

Responsibilities

• Build multi-agent workflows with guardrails and human-in-the-loop controls.
• Design retrieval pipelines: chunking, hybrid search, reranking, and citations.
• Own evaluation: golden datasets and regression suites in CI.
• Collaborate directly with client engineering teams from design to production.

Skills

Python

LLM/ML systems

Observability

Human-in-the-loop

Distributed systems

Tools

Vector stores

Inference serving

Azure

AWS

Job description

About the Role

Design and ship production agentic AI systems — multi-agent orchestration, RAG pipelines and evaluation harnesses — for enterprise clients and our BidsInsight platform.

What You'll Do

• Build multi-agent workflows with guardrails, observability and human-in-the-loop controls
• Design retrieval pipelines: chunking, hybrid search, reranking, citation-grounded generation
• Own evaluation: golden datasets, faithfulness metrics, regression suites in CI
• Work directly with client engineering teams from design to production

What We Look For

• 4+ years software engineering, 2+ years shipping LLM/ML systems to production
• Strong Python; experience with vector stores, inference serving (vLLM-class) and cloud (Azure/AWS)
• Bias for measurement — you don't ship what you can't evaluate