LLM Systems / AI Agent Engineer

Accountabilities:

• • Build, deploy, and continuously improve production AI agents using foundation models, including AWS Bedrock and comparable providers.
  • Design and evolve orchestration systems supporting tool calling, streaming, context management, structured outputs, and agent workflows.
  • Apply context-engineering techniques to improve the reliability, quality, and effectiveness of LLM and agent systems.
  • Develop and maintain evaluation datasets and pipelines covering tool selection, agent trajectories, and LLM-as-judge evaluations.
  • Establish production observability, monitoring, tracing, and instrumentation for LLM and agent workloads.
  • Contribute architectural expertise to evaluate the existing custom orchestration layer and determine whether a production framework such as LangGraph or LangChain would provide additional value.
  • Work closely with the AI function lead and Full Stack Engineer as part of a focused three-person product team.
  • Operate as an individual contributor, taking ownership of clearly defined and measurable deliverables from onboarding onward.
  • Identify production failure modes, implement appropriate mitigations, and contribute across the AI engineering stack.
  • Optimize LLM operating costs through prompt caching, model selection, routing strategies, and broader LLM FinOps practices.
  • Help establish robust engineering practices for reliable, scalable, and maintainable agentic AI systems.

  Requirements:

  • 2+ years of professional experience building production LLM agents, including tool-calling loops, streaming, context management, structured outputs, and orchestration.
  • Production experience with an agentic framework such as LangGraph, LangChain, or a custom orchestration solution, combined with strong understanding of agent architecture patterns.
  • 1.5+ years of experience with evaluation-driven development, including evaluation datasets and pipelines for tool selection, trajectory assessment, and LLM-as-judge approaches.
  • 1+ year of hands-on experience with LLM observability, tracing, and instrumentation using tools such as Langfuse, OpenTelemetry, or equivalent platforms; direct production experience with agent observability is particularly valuable.
  • 1+ year of experience optimizing LLM costs through prompt caching, model selection and routing, and LLM FinOps.
  • 5+ years of backend engineering experience, including strong proficiency with TypeScript/Node.js, PostgreSQL, and serverless AWS technologies.
  • Proven track record of shipping agentic AI systems to production and delivering comparable engineering projects against defined timelines.
  • Ability to discuss real-world agent failure modes, reliability challenges, and mitigation strategies with strong technical depth.
  • Ability to operate independently across the AI stack while collaborating effectively within a small, specialized engineering team.
  • Experience with AWS Bedrock is an advantage, while production experience with other foundation-model providers such as OpenAI, Anthropic, Azure OpenAI, or Vertex AI is transferable.
  • Knowledge of AI safety and guardrails, including prompt-injection protection, output validation, and handling untrusted inputs, is a plus.
  • Familiarity with MCP and multi-agent architectures is beneficial.
  • Exposure to geospatial data is an additional advantage.

  Benefits:

  • Fully remote, full-time working arrangement in India.
  • Fixed working hours: 12:00 PM–9:30 PM IST during summer and 1:00 PM–10:30 PM IST during winter.
  • No weekend work, supporting a strong work-life balance.
  • Laptop provided from day one.
  • Full medical insurance from the start of employment.
  • Access to mentorship, professional communities, and knowledge-sharing forums.
  • Supportive, collaborative environment focused on continuous learning and professional development.
  • Opportunity to take meaningful ownership within a small, specialized AI product team.
  • Long-term career opportunity where individual contributions and technical impact are valued.