Remote Senior AI Agent Engineer – Python & LLM Systems

Michael Page; United States; Remote; USD 67,000 - 135,000; Full time

Job summary

Michael Page is seeking a Senior AI Agent Engineer with expert Python skills to build and operate production LLM-powered agent systems. You will work with AWS Bedrock, LangGraph, and LangChain in a multinational setting and contribute to RAG pipelines, embeddings, and advanced observability.

You will drive multilingual architecture across the full pipeline, including translation accuracy and cross-lingual retrieval, while ensuring robust evaluation and latency-aware design in a remote

Qualifications

• Hands-on experience with LLM-powered agent systems in production.
• Expert Python skills and practical AI platform building.
• Experience with AWS Bedrock and agent frameworks (LangGraph, LangChain).
• Knowledge of RAG architectures embeddings vector search and observability.

Responsibilities

• Redesign state machine for multi-turn conversations with anchoring patterns.
• Refactor reasoning pipeline and routing; remove advance-only defaults.
• Improve retrieval quality via embedding pipeline cleanup and hybrid search.
• Implement multilingual solution across the full pipeline including translation and cross-lingual retrieval.
• Build and maintain offline agent evaluation framework and quality gates.
• Work on real-time speech-to-text and mishear recovery for agent layer.
• Instrument traces for latency cost per turn tool selection and session quality.

Skills

Python

LLM-powered agents

Multilingual/ cross-lingual

Observability

Tools

AWS Bedrock

LangGraph

LangChain

Strands

Job description

Michael Page is seeking a Senior AI Agent Engineer with expert Python skills to build and operate production LLM-powered agent systems. You will work with AWS Bedrock, LangGraph, and LangChain in a multinational setting and contribute to RAG pipelines, embeddings, and advanced observability.

You will drive multilingual architecture across the full pipeline, including translation accuracy and cross-lingual retrieval, while ensuring robust evaluation and latency-aware design in a remote