Job description

Experience-12–18 years overall, including 4+ years designing and delivering knowledge graph or graph-backed AI systems in production Function-AI / Data Architecture — client-facing delivery and pre-sales solutioning Location-- Onsite Role Summary We are hiring a Lead AI Architect to own knowledge graph architecture across client engagements. This is a hands-on architecture role, not a management one: you will design the ontology, choose the graph and retrieval stack, prove it on real client data, and stay close enough to the implementation to be accountable for whether it works in production. Most of our clients arrive with the same problem in different clothing. They have deployed a retrieval-augmented chatbot over a document store, it demos well, and it fails the moment a question requires reasoning across entities, hierarchies, or relationships that no single document states. Your job is to design the semantic layer that fixes that — entity models, relationship graphs, and graph-grounded retrieval that gives an LLM structured context instead of nearest-neighbour text. You will also help build the practice: reusable accelerators, reference architectures, delivery patterns, and the technical narrative our pre-sales teams take to prospects. What You Will Do Architecture and design Own the knowledge graph architecture end to end on assigned engagements — conceptual model, ontology or schema, graph data model, ingestion and enrichment pipelines, query and serving layer. Design entity and relationship models from messy enterprise reality: multiple source systems, inconsistent identifiers, partial records, and business definitions that differ by department. Define the entity resolution strategy — blocking, matching, survivorship, and the human-in-the-loop path for the cases the algorithm should not decide alone. Architect GraphRAG and graph-grounded retrieval — hybrid vector plus graph traversal, subgraph retrieval, context assembly, and prompt grounding strategies that keep answers traceable to graph facts. Make and defend build-versus-buy calls across graph databases, vector stores, ontology tooling, and orchestration frameworks, with a written rationale a client architect can challenge. Set the non-functional bar — query latency, graph scale, incremental update patterns, lineage, access control at node and relationship level, and how the graph is versioned as the ontology evolves. Delivery Lead technical discovery: work with client SMEs to extract the domain model that exists in their heads and nowhere in their systems. Build proofs of concept yourself. At this level we expect you to be able to open an editor and demonstrate the pattern, not only describe it. Guide engineering teams through implementation — design reviews, Cypher and pipeline reviews, unblocking hard modelling questions. Own the path from pilot to production: evaluation harnesses, monitoring, drift and quality checks on graph content, and clear operational ownership before go-live. Set graph quality metrics and hold the delivery to them — coverage, resolution precision and recall, answer groundedness, and retrieval relevance. Practice and pre-sales Build reusable accelerators, reference architectures, and delivery playbooks for graph and GraphRAG engagements. Support pre-sales: solution shaping, effort and cost estimation, architecture sections of proposals, and technical defence in client evaluations. Mentor architects and senior engineers into graph modelling competence — this capability cannot rest on one person. Represent the practice externally where useful: client architecture forums, partner sessions, conference talks, published reference material. Required Experience 12–18 years in software, data, or AI engineering, with a clear architecture track record on systems that reached production and stayed there. 4+ years hands-on with knowledge graphs — you have designed the model, not only queried someone else’s. Deep property graph expertise — Neo4j in production, fluent Cypher including query tuning, and sound judgement on graph data modelling trade-offs (supernodes, relationship direction and cardinality, index and constraint strategy, projection patterns). Demonstrated GraphRAG or graph-plus-LLM delivery — hybrid retrieval, subgraph extraction and context assembly, grounding and citation strategy, and honest evaluation of whether the graph actually improved answer quality. Ontology and semantic modelling — taxonomies, hierarchies, controlled vocabularies, and the discipline to keep a model coherent as scope grows. Entity resolution and data quality at scale — practical experience with the failure modes, not just the theory. LLM application engineering — orchestration frameworks, embedding and chunking strategy, evaluation methodology, cost and latency management, and the limits of what prompting can fix. Strong data engineering foundation — Python, SQL, distributed processing, streaming or batch ingestion, and modern pipeline tooling. Cloud architecture on at least one major platform (Azure preferred), including managed graph, vector, and AI services. Client-facing credibility — you can hold an architecture conversation with a client CTO and a modelling conversation with their engineers in the same afternoon. Preferred Experience RDF, SPARQL, OWL, or SHACL, and experience with triple stores such as GraphDB, Stardog, or Amazon Neptune — useful when clients arrive with standards-based estates. Other graph platforms: TigerGraph, Amazon Neptune, Azure Cosmos DB Gremlin API, Memgraph. Graph algorithms and graph machine learning — centrality, community detection, link prediction, graph embeddings, GNNs. Agentic architectures where a graph provides tool grounding, memory, or planning context. Domain depth in a regulated industry — healthcare and life sciences, financial services, manufacturing, or the public sector. Master data management, data cataloguing, or data governance background. Consulting or systems integrator experience with multi-client delivery. Relevant certifications: Neo4j Certified Professional or Graph Data Science, Azure Solutions Architect Expert, Azure AI Engineer. What We Look For Beyond the stack, the architects who succeed here tend to share a few traits: They model the domain before they choose the tool, and they can explain why a graph is the right answer — or say plainly when it is not. They are sceptical of demos, including their own, and insist on evaluation before claiming a result. They write things down. Decisions, trade-offs, and rejected options, in a form the client still understands six months later. They design for the team that inherits the system, not for the elegance of the diagram. They are comfortable saying "this pilot should not go to production yet" to a client who wants to hear otherwise. First-Year Success Measures First 90 days — fluent in our current engagements and accelerators; owning knowledge graph architecture on at least one active client programme; a documented view of where our practice capability is thin. Six months — one graph or GraphRAG solution delivered into production with an evaluation harness and named operational ownership; at least one reusable accelerator or reference architecture contributed. Twelve months — graph architecture patterns adopted across multiple accounts; two or more architects independently competent in graph modelling; measurable pre-sales contribution to won work. Education Bachelor’s degree in Computer Science, Engineering, or a related field. A Master’s or PhD in computer science, computational linguistics, semantic technologies, or a related discipline is welcome but not required — demonstrated production delivery carries more weight.

Requirements

Must-haves

Experience-12–18 years overall, including 4+ years designing and delivering knowledge graph or graph-backed AI systems in production
Function-AI / Data Architecture — client-facing delivery and pre-sales solutioning
Location-- Onsite
You will also help build the practice: reusable accelerators, reference architectures, delivery patterns, and the technical narrative our pre-sales teams take to prospects.
Architecture and design
Own the knowledge graph architecture end to end on assigned engagements — conceptual model, ontology or schema, graph data model, ingestion and enrichment pipelines, query and serving layer.
Design entity and relationship models from messy enterprise reality: multiple source systems, inconsistent identifiers, partial records, and business definitions that differ by department.
Define the entity resolution strategy — blocking, matching, survivorship, and the human-in-the-loop path for the cases the algorithm should not decide alone.
Architect GraphRAG and graph-grounded retrieval — hybrid vector plus graph traversal, subgraph retrieval, context assembly, and prompt grounding strategies that keep answers traceable to graph facts.
Make and defend build-versus-buy calls across graph databases, vector stores, ontology tooling, and orchestration frameworks, with a written rationale a client architect can challenge.
Set the non-functional bar — query latency, graph scale, incremental update patterns, lineage, access control at node and relationship level, and how the graph is versioned as the ontology evolves.
Lead technical discovery: work with client SMEs to extract the domain model that exists in their heads and nowhere in their systems.

Nice-to-haves

Cloud architecture on at least one major platform (Azure preferred), including managed graph, vector, and AI services.
RDF, SPARQL, OWL, or SHACL, and experience with triple stores such as GraphDB, Stardog, or Amazon Neptune — useful when clients arrive with standards-based estates.
Other graph platforms: TigerGraph, Amazon Neptune, Azure Cosmos DB Gremlin API, Memgraph.
Graph algorithms and graph machine learning — centrality, community detection, link prediction, graph embeddings, GNNs.
Agentic architectures where a graph provides tool grounding, memory, or planning context.
Domain depth in a regulated industry — healthcare and life sciences, financial services, manufacturing, or the public sector.
Master data management, data cataloguing, or data governance background.
Consulting or systems integrator experience with multi-client delivery.
Relevant certifications: Neo4j Certified Professional or Graph Data Science, Azure Solutions Architect Expert, Azure AI Engineer.
What We Look For
Beyond the stack, the architects who succeed here tend to share a few traits:
They model the domain before they choose the tool, and they can explain why a graph is the right answer — or say plainly when it is not.

Application prep

Use the role’s language and application details while tailoring your materials.

Keywords to mirror

aillmgorestpythonsqlazureapimachine learning

Platform: careers-browser
Effort: medium