Forward Deployed Engineer

Insight Global; Hyderabad; On-site; INR 3,500,000 - 6,500,000; Full time

Job summary

Insight Global in India seeks a Forward Deployed Engineer to build agentic AI systems for enterprise clients, delivering production-ready solutions. You will own end-to-end delivery within client engagements, collaborate with Technical Architects and POD engineers, and help create reusable IP.

The role emphasizes reliability, observability, and practical impact in real business processes. This is frontline backbone engineering: shaping architecture, tuning models, and ensuring systems stay

Qualifications

• 8+ years building ML and AI systems in production.
• Strong software and ML engineering experience.
• Experience with agentic systems, RAG, and orchestration.

Responsibilities

• Tackle the hardest model and agent problems in client projects with reliability.
• Design agentic workflows: LLM orchestration, multi-agent systems, tooling calls.
• Prototype and prove approaches with measurable, production-ready results.
• Architectural input with Technical Architects to set reusable patterns.
• Build production-grade RAG, retrieval, prompts, and agent runtimes.
• Implement MLOps: deployment, monitoring, drift detection, health checks.
• Deliver end-to-end in live client engagements and enable reuse.

Skills

ML Engineering

Python

Production Systems

Software Engineering

Tools

LangGraph

LangChain

Eval Harnesses

LangSmith

AWS Bedrock

SageMaker

Azure OpenAI

GCP Vertex AI

Python

Job description

We are looking for a Forward Deployed Engineer to build the agentic systems at the core of what IG Labs delivers. Our PODs deploy AI agents into enterprise clients to run real business processes, and you are the engineer who builds them: the models, the orchestration, the agent logic, and the evaluation that make a system good enough for a client to put into production. This is backbone engineering, the work the whole organization is built around, not a function that supports it from the side.

You take on the hardest AI problems in our client work: the agent that has to be reliable enough to trust, the retrieval that has to be accurate over messy enterprise data, the evaluation that proves the system is good before a client stakes a process on it. You own the AI that goes live in a client environment, from a rough problem to a system in production, and you are accountable for whether it actually works.

Everything you build, you build to last beyond one client. The agents, components, and evaluation frameworks you create become Factory IP that compounds across every engagement, so each project starts ahead of the last. You are a senior engineer in our delivery organization, based in India, working within the POD model alongside the Technical Architects and the Data, UI/UX, and Technical/Field engineers who build the rest of the system.

Key Responsibilities:

Solving the Hardest AI Problems

• Take on the toughest model and agent problems in our client work: accuracy, reliability, and the behaviors a client will actually stake a process on.
• Design and build agentic workflows: LLM orchestration, multi-agent systems, tool use and function calling, and the human-in-the-loop boundaries that make autonomy safe.
• Take an open-ended AI problem with no established pattern and prototype, measure, and prove an approach that holds up.
• Own the question of whether the system is good: design the evaluation harnesses, guardrails, and metrics that separate a demo from something production-ready.
• Shape genuinely architectural decisions with the Technical Architect, contributing the patterns that become standard.

Agentic & ML Engineering

• Build production-grade RAG and retrieval, prompt and context engineering, and agent runtimes, with deep, current command of the modern LLM stack.
• Make the hard model tradeoffs across selection, accuracy, latency, and cost, tuned to what the engagement actually needs.
• Fine-tune or adapt models where it earns its keep, and know when it does not.
• Stand up the MLOps that keep agentic systems honest in production: deployment, monitoring of model and agent health, drift and regression detection.
• Build the evaluation and observability that prove value to a client with evidence, not assertion.

Delivering in Client Engagements

• Own your work end to end within live client engagements, from a rough problem to an agentic system running in production.
• Bring the deep AI and agentic expertise a client engagement depends on, as a peer to the Technical Architects, Forward Deployed Engineers, and other engineers in the POD.
• Maintain the documentation and handoff discipline that distributed, multi-client delivery requires, so your work is easy to pick up, extend, and reuse.
• Be available for the hard, real-time problems that need a working session, not a write-up.

Building Reusable Capability

• Turn everything you solve into reusable capability: agent components, retrieval patterns, evaluation frameworks, and prompts that make the next engagement faster.
• Partner with the Technical Architect to abstract, validate, and document what you build for the IP library.
• Actively drive Reuse Index. The agentic building blocks you create are how the next project starts ahead of where the last one finished.
• Treat capability-building as core, not overhead. Every hard problem solved should leave the Factory stronger.

What This Role Is Not

• Not a research role. You are not here to write papers or perfect notebooks. You ship agentic systems into live client production.
• Not a prompt-tweaker. You engineer AI systems end to end: retrieval, evaluation, orchestration, and deployment.
• Not an internal platform team. You build the systems clients pay for, not internal tooling for its own sake.
• Not the Technical/Field Engineer. They master the client's stack and prove feasibility in it; you build the AI and agentic systems that run on it. Different seats, both core.
• Not the Technical Architect. The TA sets architecture and standards across engagements; you build the systems within them and shape the patterns that become standards.

Qualifications

Required

Strongly Preferred

8+ years building machine learning and AI systems in production, with recent, hands-on depth in LLM and agentic systems: RAG, orchestration, multi-agent design, and evaluation. Strong software and ML engineering in Python. A track record of taking models and agents to production and keeping them reliable (MLOps, monitoring, evaluation). Methodical about accuracy and proving a system is good. A clear written communicator who collaborates well across a distributed team.

Depth with modern agent frameworks and orchestration, fine-tuning and model adaptation, and vector stores and retrieval systems. Experience in a consulting, professional services, or client-delivery environment. Familiarity with the data, access, and compliance constraints of regulated clients in financial services or healthcare. A history of building capability that compounds across a wider team.

Tools & Technology

Languages - Python

LLM & Agentic -LLM orchestration, agent frameworks (LangGraph, LangChain), tool use and function calling

Evaluation & Guardrails- Eval harnesses, LLM observability (LangSmith), guardrails

Cloud AI Platforms - AWS Bedrock and SageMaker, Azure OpenAI, GCP Vertex AI