Generative AI Engineer - Knowledge Graph

Description :

Knowledge Graph Engineering :

- Design, build and maintain large-scale property graphs and RDF triplestores (Neo4j, Amazon Neptune, Stardog, or equivalent).

- Develop and govern ontologies, taxonomies, and entity-relationship schemas that reflect real-world domain semantics.

- Implement graph ingestion pipelines that extract, transform, and link entities from structured, semi-structured, and unstructured data.

- Optimise graph traversal queries (Cypher, SPARQL, Gremlin) for sub-second response at production scale.

- Train and deploy graph neural networks (GNNs) for node classification, link prediction, and subgraph retrieval - Maintain model retraining workflows triggered by graph drift or coverage degradation.

Agentic AI Systems :

- Architect and implement autonomous agents that plan multi-step reasoning chains over knowledge graph data using LLMs (GPT-4o, Claude, Gemini, or open-source equivalents).

- Build graph-aware Retrieval-Augmented Generation (RAG) pipelines that blend structured graph context with unstructured document retrieval.

- Design tool-use and function-calling layers so agents can query live data sources web search, REST/GraphQL APIs, relational databases to extend or verify graph knowledge.

- Implement agent memory, reflection, and self-correction loops to improve reliability over multi-hop tasks.

Context Enrichment & Data Fusion :

- Integrate web scraping, news feeds, and open-source intelligence (OSINT) sources to keep the knowledge graph current.

- Build entity resolution and deduplication components that merge data from heterogeneous sources into a consistent graph.

- Develop confidence-scoring and provenance-tracking mechanisms so downstream consumers understand the reliability of any piece of context.

MLOps & Production Readiness :

- Package agents as scalable microservices; instruments with observability tooling (tracing, latency, token cost).

- Collaborate with platform engineers to deploy workloads on cloud-native infrastructure (AWS / GCP / Azure).

- Maintain evaluation harnesses that measure agent accuracy, hallucination rate, and graph coverage over time.

- 5 + years of professional software engineering with strong Python (or Java / Kotlin) proficiency.

- Hands-on production experience with at least one major graph database Neo4j, Amazon Neptune, TigerGraph, or comparable.

- Demonstrated knowledge of graph query languages like Cypher, SPARQL, or Gremlin at production query complexity.

- Direct experience building LLM-powered agents or pipelines using frameworks such as LangChain, LangGraph, LlamaIndex, CrewAI, AutoGen, or Semantic Kernel.

- Solid understanding of RAG architectures: chunking strategies, vector stores (Pinecone, Weaviate, pgvector), hybrid retrieval, and re-ranking.

- Familiarity with prompt engineering, few-shot learning, and LLM evaluation techniques.

- Experience integrating external data sources via APIs, web scraping (Playwright / Scrapy), or streaming pipelines (Kafka / Kinesis).

- Working knowledge of containerisation (Docker, Kubernetes) and CI/CD pipelines.

- Familiarity with graph export formats - at least one GraphML, RDF/OWL, or JSON-LD

- Experience integrating GNN-derived features into vector stores or RAG pipelines

- Advanced degree (MS / PhD) in Computer Science, Information Science, Computational Linguistics, or a related field.

- Experience in intelligence, defence, or trade-craft environments working with OSINT, link analysis, entity disambiguation, or signals intelligence data.

- Understanding of access-control models for sensitive graph data (need-to-know, compartmentalisation, provenance labelling).

- Familiarity with knowledge representation standards like OWL, SHACL, RDF-star, JSON-LD, W3C PROV.

- Experience with fine-tuning or instruction-tuning open-source LLMs (Llama, Mistral, Falcon) for domain-specific tasks.

- Background in network-analysis algorithms: centrality, community detection, path-finding, anomaly detection on graphs.

- Contributions to open-source graph or GenAI projects; published research or technical blog presence.

- Active or adjudicatable security clearance (Secret or above) strongly preferred for trade-craft assignments.