AI Systems Architect – LangGraph Framework & RAG Pipelines

Job Overview We are seeking an experienced AI Engineer with deep expertise in building agentic AI systems, complex multi-agent graphs, and high-performance RAG pipelines. In this role, you will design, build, and optimize autonomous agent workflows using LangGraph, integrate vector databases for retrieval-augmented generation, and connect AI agents to external APIs and dynamic toolsets. Key Responsibilities Agentic System Architecture: Design and implement multi-agent workflows, state management, and fallback mechanisms using LangGraph (or similar orchestration frameworks). Vector DB & RAG Pipeline Development: Set up, index, and optimize vector databases (e.g., Pinecone, Qdrant, Chroma, Weaviate, or pgvector) for fast, context-aware retrieval. Tool & API Integration: Build dynamic function-calling pipelines and custom agent tools to execute real-world tasks. Performance & Reliability Tuning: Minimize latency, manage context windows efficiently, and implement robust error-handling/retry logic for non-deterministic model outputs. Evaluation & Benchmarking: Establish evaluation metrics (e.g., using Ragas, TruLens, or custom benchmarks) to validate accuracy, hallucination rates, and execution paths. Required Skills & Experience Frameworks: Expert knowledge of LangGraph, LangChain, or similar agent frameworks. Vector Databases: Hands-on experience with vector search, hybrid retrieval (dense + sparse), metadata filtering, and embedding model selection. Programming & Tech Stack: High proficiency in Python or TypeScript, async programming, and API development (FastAPI/Express). LLM Architecture: Deep understanding of function calling, structured outputs, prompt engineering, context window management, and agent memory state persistence. Data Pipelines: Experience processing, chunking, and embedding unstructured documents (PDFs, JSON, web content). Preferred Qualifications Experience deploying agentic applications to production environments (AWS, GCP, Modal, Vercel). Familiarity with human-in-the-loop (HITL) workflows within LangGraph state machines. Background in observability platforms like LangSmith or Phoenix. Briefly describe a complex multi-agent system or LangGraph workflow you have built and deployed to production.