Agentic AI Developer

Xpetize; Thiruvananthapuram; Hybrid; INR 1,800,000 - 2,800,000; Full time

Job summary

Xpetize is seeking an Agentic AI Developer in India to design and ship AI agents at the core of Synthetix. You will engineer reasoning loops, tool use, memory, and governance rails for trustworthy, observable agents that run in production.

You’ll prototype quickly, harden what works, and own agents from design through production, collaborating with platform, product, and design teams to optimize cost and latency across OpenAI, Claude, Gemini, and open‑weight models.

Qualifications

• Solid production software experience with hands-on AI agents/LLM apps.
• Proficiency in Python and TypeScript/JavaScript for full-stack integration.
• Hands-on with at least one agent framework and one orchestration framework.
• Strong grasp of prompting, tool calling, structured outputs and context management.

Responsibilities

• Design, build, and ship AI agents using modern frameworks and pick the right tool for each problem.
• Architect multi-agent orchestration and stateful workflows with LangGraph and LangChain.
• Enable agent-to-agent communication for collaboration across systems.
• Implement robust tool calling, MCP servers, and connectors for external systems.
• Build RAG pipelines, embeddings, vector search, and agent memory.
• Incorporate governance, guardrails, human-in-the-loop checkpoints, and audit trails.
• Develop end-to-end evals, datasets, LLM-as-judge scoring, and regression tracking.
• Integrate agents into full-stack apps across APIs, services, and front-ends.
• Optimize cost, latency, and quality across providers like OpenAI, Claude, Gemini.

Skills

Python

TypeScript/JavaScript

LLM prompts

RAG/embeddings

REST APIs

Communication

Agent frameworks

orchestration

Debugging

Cloud awareness

Tools

OpenAI Agents SDK

Google ADK

LangChain/LangGraph

AutoGen

CrewAI

Job description

Agentic AI Developer

The Role

We're looking for an Agentic AI Developer who lives at the intersection of applied LLM engineering and production software. You'll design and ship the AI agents and orchestration flows at the core of Synthetix: reasoning loops, tool use, multi-agent coordination, memory, and the governance rails that keep them trustworthy.

You're comfortable across the stack, but your center of gravity is agent creation: turning a capability into a reliable, observable, evaluable agent. This is a hands-on builder role. You'll prototype fast, harden what works, and own agents from design through production.

What You'll Do

• Design, build, and ship AI agents using modern agent frameworks (OpenAI Agents SDK, Google's Agent Development Kit (ADK), Microsoft Agent Framework / AutoGen, CrewAI, or equivalents), picking the right tool for each problem.
• Architect multi-agent orchestration and stateful workflows with LangGraph, LangChain, and similar frameworks: planning, routing, hand-offs, retries, and durable long-running execution.
• Enable agents to talk to each other reliably, using agent-to-agent (A2A) communication patterns and protocols so agents can share context, delegate work, and collaborate across systems.
• Implement robust tool / function calling, structured outputs, and external-system integration, including Model Context Protocol (MCP) servers and connectors.
• Build the retrieval and memory layers: RAG pipelines, embeddings, vector search, and short- and long-term agent memory.
• Bake in governance from day one: guardrails, policy enforcement, human-in-the-loop checkpoints, audit trails, and safe-failure behavior. This is the "governed" in governed orchestration.
• Build and own the evals system end to end: test datasets, automated evals that run on every change, LLM-as-judge scoring, and regression tracking, backed by tracing and telemetry (e.g., LangSmith, Langfuse, OpenTelemetry) to measure and continuously improve agent reliability.
• Integrate agents into full-stack applications across APIs, services, and front-end surfaces, collaborating closely with platform, product, and design.
• Optimize for cost, latency, and quality across model providers (OpenAI, Anthropic / Claude, Google / Gemini, and open-weight models).

What You'll Bring

• 3+ years building and shipping production software, with meaningful recent hands-on experience building AI agents or LLM-powered applications.
• Strong Python (the working language of agent development) and solid TypeScript / JavaScript for full-stack integration.
• Hands-on experience with at least one agent framework / SDK (OpenAI Agents SDK, Google ADK, AutoGen, CrewAI, or similar) and one orchestration framework (LangChain / LangGraph or equivalent).
• A solid grasp of LLM building blocks: prompting, function / tool calling, structured outputs, context-window management, RAG, and embeddings / vector databases.
• A good understanding of agent harnesses, the layer around the model that runs the agent loop, executes tools, manages context, and controls permissions.
• Familiarity with Spec-Driven Development (SDD) and AI-driven development methods like BMAD, and how to use them to build with AI coding agents in a structured way.
• Solid understanding of the software development lifecycle (SDLC), from requirements and design through testing, deployment, and maintenance.
• Full-stack fluency: REST / async APIs (FastAPI, NestJS, or similar), relational databases (PostgreSQL), and enough front-end (React) to build and integrate end-to-end.
• Familiarity with cloud (AWS, Azure, or GCP), containers (Docker), and CI/CD.
• A product mindset, strong debugging instincts, and the judgment to make non-deterministic systems behave reliably.
• Strong communication skills, with the ability to explain technical ideas clearly to both technical and non-technical audiences.

Nice to Have

• Experience with agent evaluation and LLM observability tooling (LangSmith, Langfuse, Arize, and similar).
• Experience with MCP, Temporal or other durable-execution engines, and event-driven architectures.
• Exposure to enterprise / regulated environments: SSO, RBAC, data governance, and compliance.
• Insurance domain experience, such as underwriting, claims, or policy servicing etc.
• Fine-tuning, prompt-optimization, or model-routing experience.
• Open-source contributions in the agent / LLM ecosystem.