Senior AI Engineer (Node.js & LangGraph / RAG)

We are looking for a backend-focused AI Engineer experienced in Node.js and LangGraph to refine and scale our production RAG pipeline. Our system processes incoming emails, classifies intent, extracts and parses attached PDFs, and powers a high-precision question-answering interface. Your core mission is to optimize this architecture—built on Node.js and LangGraph—to consistently achieve 95% end-to-end system accuracy. Goal: 95% Pipeline Accuracy & Deep Retrieval Testing Your primary KPI will be lifting and maintaining 95% or more accuracy across multi-stage ingestion and retrieval. Because the pipeline handles personal data spanning emails and deep-ingested PDFs, it must pass rigorous evaluation tests for complex, time-sensitive, and multi-entity user queries, such as: "What are my upcoming reminders?" "What are top priorities next week?" Key Accuracy Benchmarks & Responsibilities LangGraph Pipeline Refinement: Design, debug, and optimize stateful multi-agent workflows and graph-based routing in LangGraph (Node.js) to handle email classification, PDF extraction parsing loops, and fallback mechanisms. Complex Temporal & Contextual QA: Fine-tune chunking, metadata tagging, and retrieval strategies so the RAG layer accurately parses dates, schedules, and priority markers across nested documents without hallucinations or missed context. End-to-End Evaluation Harnesses: Build automated test suites within the Node.js ecosystem to continuously measure retrieval precision, recall, and answer faithfulness against complex prompt datasets. Document Extraction & Ingestion Hardening: Eliminate noise and parsing loss when extracting and structuring text/tables from diverse PDF attachments. Key Responsibilities LangGraph Orchestration: Own the state management and execution flows of our ingestion and retrieval graphs using LangGraph in Node.js. Retrieval & Embedding Optimization: Optimize vector search, hybrid keyword/semantic search, and re-ranking mechanisms to maximize context relevance. Classification Tuning: Refine classification models or LLM router nodes within the graph to accurately categorize incoming emails and attachments. Performance & Scalability: Ensure asynchronous email ingestion workers, PDF parsers, and LangGraph execution steps run efficiently in a Node.js cloud environment. Requirements & Qualifications Experience: 3+ years of software engineering experience, with specialized expertise in building production-grade LLM applications, RAG pipelines, and agentic workflows. Core Technical Stack: Languages: Node.js / JavaScript / TypeScript (Primary backend environment). AI Orchestration: LangGraph.js, LangChain.js, or equivalent agentic graph frameworks. LLMs & Vector DBs: OpenAI APIs / open-source LLMs, Vector Databases (e.g., Pinecone, Qdrant, Milvus, pgvector). Document Processing: Experience with PDF parsers, OCR, and document structure extraction tools. Testing & Accuracy Mindset: Proven ability to build evaluation benchmarks, debug complex graph state issues, and systematically drive LLM pipeline accuracy above 95%.