AI Data Engineer III

About the role

Rimini Street is hiring an AI Data Engineer III for a remote position based in Brazil. The engineer builds the knowledge layer of the Agentic ERP platform: data pipelines, RAG (retrieval-augmented generation) systems and embedding infrastructure that let AI agents access the right information at the right time. The role reports to the Sr. Director, Engineering.

Key responsibilities

• Design and build RAG pipelines, including chunking strategies for support tickets, documentation, policies, transaction records and email threads, and hybrid retrieval combining dense embeddings, sparse search (BM25) and metadata filtering.
• Implement vector storage with PostgreSQL pgvector, evaluate embedding models, build embedding pipelines and design multi-tenant vector architectures.
• Build data pipelines that ingest knowledge from Salesforce, ServiceNow, documentation repositories, email archives and ERP logs, with ETL, data quality monitoring and lineage tracking.
• Design the knowledge architecture, including knowledge graphs, metadata taxonomies, versioning and feedback loops.
• Integrate with Snowflake, moving and transforming data between operational systems, Snowflake and vector stores while respecting customer data isolation.

Required profile

• 5+ years of data engineering experience, with at least 1-2 years on AI/ML pipelines or RAG systems.
• Hands-on experience building and optimizing RAG pipelines in production.
• Bachelor's or Master's degree in Computer Science, Data Science or a related field preferred.
• Fluent English, written and spoken; self-motivated in a remote environment.

Required skills

• Python for data engineering (pandas, async programming)
• PostgreSQL with strong SQL (JSONB, full-text search, extensions)
• pgvector or similar vector databases (Pinecone, Qdrant, Weaviate)
• ETL/ELT and pipeline orchestration (Airflow, Dagster, Prefect)
• Git, CI/CD and REST/GraphQL API integration
• Preferred: Snowflake, LlamaIndex/LangChain/Haystack, dbt, Neo4j, Elasticsearch/OpenSearch