Agentic AI Lead (Python) — Vertex AI RAG + GraphVector Datastores

Job Title

Agentic AI Lead (Python) - Vertex AI RAG Graph/Vector Datastores

Berkeley Heights NJ (5 Days Onsite)

FTE

Senior candidates only with min 10 years of experience

Role summary

Were looking for a strong agentic AI developer who can build and productionize Vertex AI based RAG systems (Vertex AI Search / Vertex AI RAG patterns) design reliable tool-using agents and work comfortably with vector databases and graph databases. Youll own end-to-end delivery: ingestion retrieval agent orchestration evaluation deployment.

What youll do

• Design and implement RAG pipelines on Google Cloud / Vertex AI (chunking embeddings indexing retrieval reranking grounding).
• Build agentic workflows (tool use planning reflection/guardrails structured outputs) using Python-first frameworks.
• Integrate agents with Graph DBs (e.g. Neo4j JanusGraph Neptune) and Vector DBs (e.g. Vertex Vector Search Pinecone Weaviate Milvus pgvector).
• Create robust data ingestion/ETL from PDFs docs webpages and internal sources; implement metadata strategy and access control.
• Define and run evaluation (retrieval metrics answer quality hallucination/grounding checks) and improve system quality iteratively.
• Ship to production: APIs monitoring/observability cost/performance optimization CI/CD and security best practices.

Must-have skills

• Strong Python (clean architecture async testing typing packaging).
• Proven experience building RAG solutions (hybrid search reranking chunking strategies embeddings prompt schema design).
• Hands-on with Vertex AI and GCP fundamentals (IAM logging/monitoring Cloud Run/GKE storage).
• Experience with at least one agentic framework (e.g. LangGraph/LangChain LlamaIndex Semantic Kernel AutoGen) and tool/function calling patterns.
• Solid knowledge of vector search concepts and at least one vector DB in production.
• Comfortable with graph data modeling and graph querying (Cypher/Gremlin/SPARQL basics).
• Strong engineering practices: code reviews testing telemetry secure-by-design reliability mindset.

Nice-to-have

• Knowledge graphs for RAG (entity linking graph traversal retrieval fusion).
• Streaming/messaging (Pub/Sub Kafka) document pipelines (Document AI) and multilingual retrieval.
• Experience with evaluation tooling (RAGAS TruLens custom eval harnesses) prompt/version management.