Graph / RAG Engineer - Remote
Experience: 7+ years, with production experience building graph and retrieval systems at scale.
You own the knowledge layer: existing homegrown code-graph generator and RAG toolchain, and the evidence-pack assembler that feeds the reasoning tier. The guiding principle of the whole design is that the model never goes hunting — you are the reason that holds true.
Responsibilities
Assess the homegrown code-graph generator — coverage, resolution percentage, correctness — and recommend build-versus-leverage; then improve it.
Own the evidence-pack assembler: deterministically gather, per item, the complete bounded context (structure, reference graph, access pattern, liveness, target-side match, business-rule linkage) so the reasoning tier never has to retrieve for itself.
Design, scale and operate the code/knowledge graph (Neo4j or Amazon Neptune): ingestion pipelines, cross-graph correlation across schema, repositories and batch, and impact-simulation queries.
Tune the retrieval pipeline — embeddings strategy, hybrid dense/sparse retrieval, and grounding-quality measurement.
Harden the RAG toolchain for production on AWS: high availability, backup and recovery, and cost-efficiency at scale.
Build the natural-language query surface over the graph ("what breaks if we retire table X").
Qualifications
Python — production-grade.
Graph databases — Neo4j and/or Amazon Neptune; Cypher / Gremlin; graph modelling and query optimisation at scale.
Code-graph and static-analysis construction: ASTs, call graphs, cross-repository symbol resolution — across multiple language stacks.
Production RAG: vector stores, embeddings strategy, hybrid retrieval, grounding evaluation.
AWS — scaling, high availability, cost-efficient operation.
Comfortable working across seven to eight different language and tech stacks.
Working knowledge
LangGraph / LangChain; ontology and semantic-layer design.
SQL Server; Informatica metadata models; CI/CD automation.
Never miss an opportunity! Create an alert based on the job you applied for.
