LLM Systems / AI Agent Engineer (Remote, Full-Time) [AS311]

Smart Working

IndiaremotePosted 16 days ago

Skill Required

LLM-Systems-EngineeringAI-Agent-EngineerBackend-EngineeringProduct-EngineeringAI-Agent-Systems-EngineerAI-Agent-Platform-EngineerAI-Agentic-Systems-EngineerApplied-AI-Agent-EngineerAI-Agents-EngineerAgent-EngineerAI-ML-EngineerAI EngineerGenerative AIAIObservabilityEngineeringTypeScriptServerlessLangChainbuildingPostgreSQLOpenAPIAzureNode.jsAWSandMachine LearningFulltime

Key highlights

First dedicated AI-agenting hire
AWS Bedrock production experience
Langfuse strongly preferred for observability
Individual contributor role, not a management position
Three-person product team structure
Fixed IST shifts with no weekend work

Role overview

Smart Working is a highly-rated remote-first workplace that connects exceptional professionals with outstanding global teams. As an LLM Systems / AI Agent Engineer, you will be the first dedicated AI-agenting hire, building and evolving production AI agents on foundation models. You'll work directly alongside the current AI function lead to establish orchestration approaches, implement evaluation pipelines, and build production observability for LLM systems. This role involves significant technical decision-making around agent architecture patterns while working as an individual contributor within a new three-person product team.

Responsibilities

Build and evolve production AI agents on foundation models, currently AWS Bedrock
Develop and evolve orchestration for production AI agents
Apply context engineering to production LLM and agent systems
Build and maintain evaluation datasets and pipelines, including tool-selection, trajectory and LLM-as-judge evaluations
Build and maintain production observability and monitoring for LLM and agent systems
Implement and work with tracing and instrumentation for production LLM systems
Work directly alongside the engineer currently leading the AI-agenting function as the first dedicated hire in this area
Apply strong agent-architecture fundamentals to help inform whether the existing custom orchestration layer should be retained or a production framework adopted
Work as part of a new three-person product team alongside the AI function lead and a Full Stack Engineer
Operate as an individual contributor working alongside the AI function lead rather than managing others
Take ownership of measurable deliverables immediately upon onboarding
Deliver similar AI/agent engineering work against roadmap timelines

Requirements

2+ years of experience building production LLM agents, including tool-calling agent loops, streaming, context management, structured outputs and orchestration frameworks
Production experience with an agentic framework such as LangGraph, LangChain or custom orchestration. There is no fixed orchestration framework requirement; strong agent-architecture fundamentals and production experience with any agentic framework are in scope
1.5+ years of experience with evaluation-driven development, including building and maintaining evaluation datasets and pipelines covering tool selection, trajectory evaluation and LLM-as-judge
1+ year of experience with LLM observability, tracing and instrumentation using Langfuse, OpenTelemetry or similar tooling
Genuine production agent-observability exposure. Direct, hands-on Langfuse experience is strongly preferred because this is a confirmed skill gap within the team; OpenTelemetry or other tracing tools are acceptable only as a secondary signal alongside real agent-observability exposure
1+ year of experience with LLM cost optimisation, including prompt caching, model selection and routing, and LLM FinOps
5+ years of backend engineering proficiency, including TypeScript/Node, Postgres and serverless AWS
Proven experience delivering similar work on similar timelines
Experience shipping agentic AI systems to production, with the ability to speak to concrete failure modes and mitigations and operate end-to-end across the AI stack

Nice to have

1+ year of AWS Bedrock experience. Equivalent production experience with other foundation-model providers, including OpenAI, Anthropic API, Azure OpenAI or Vertex AI, is fully transferable
1+ year of experience with AI safety and guardrails, including prompt-injection screening, output validation and handling untrusted input
6+ months of familiarity with MCP and multi-agent patterns
Familiarity with geospatial data