AI Agent developer – Multi-Agent system Automation & LLM Orchestration Platform (Production Systems)

We are building a production-grade AI agent system that will operate as a core automation layer across multiple internal business workflows and external integrations. This is not a prototype or chatbot-style implementation. The focus is on building reliable, stateful, tool-using agents that can execute multi-step tasks, interact with APIs, maintain context, and operate with minimal supervision in real business environments. Project Scope The system will include multiple specialized agents, each responsible for different operational domains such as: - Data processing and enrichment pipelines - API orchestration across third-party services (CRM, email, analytics, internal databases) - Automated decision-making workflows with guardrails and validation layers - Retrieval-augmented generation (RAG) over structured and unstructured data - Task execution with retry logic, logging, and audit trails - Human-in-the-loop approval flows for sensitive operations Core Requirements We are looking for someone with deep experience in: - Python backend development (FastAPI or similar frameworks) - LLM integration (OpenAI, Anthropic, or equivalent APIs) - Agent frameworks (LangChain, LangGraph, CrewAI, or custom implementations) - Tool calling / function calling architectures - Workflow orchestration and queue systems (Redis, Celery, or similar) - Database design for stateful agent memory (PostgreSQL, vector databases) - API integrations (REST, GraphQL, OAuth-based systems) What You Will Be Building - Multi-agent architecture with clear separation of responsibilities - Central orchestration layer for routing tasks between agents - Memory systems (short-term + long-term contextual storage) - Monitoring and debugging layer for agent actions and decisions - Scalable backend infrastructure for concurrent task execution - Extensible framework for adding new tools and capabilities over time Important Considerations - Production reliability is critical; hallucination handling and validation layers are required - Every agent action must be traceable and logged - System must be designed for scaling workflows, not single-use prompts - Clean, modular code architecture is expected Ideal Background - Experience building production AI systems beyond demos - Strong understanding of distributed systems and backend architecture - Experience building automation platforms or internal AI tooling - Ability to design systems that combine deterministic logic with LLM reasoning Next Step If you have built real-world AI systems involving agents, automation pipelines, or LLM-driven workflows, share relevant architecture examples or projects you’ve worked on. Focus on systems you’ve deployed, not experimental notebooks or prototypes.