Knowledge & Data Platform Lead

Knowledge & Data Platform Lead Overview

Company Name	Capgemini
Job Role	Knowledge & Data Platform Lead
Qualifications	Not Specified
Category	IT Jobs
Job Type	Full Time
Location	London

This role sits within Capgemini Financial Services as part of a new AI-native product organisation that is being built to create software products for insurance claims, payment operations, and health operations. The business is focused on product delivery rather than one-off projects, and the platform supports multiple product lines on a shared foundation. The engineering approach is deliberately senior and compact, with AI agents handling much of the implementation work while people remain responsible for the important decisions. In regulated areas, some decisions are intentionally kept human-only.

The position is centred on the knowledge and data layer that powers those AI agents. The information used by the agents is regulated, sensitive, and spread across client environments, so the work is about turning that information into something safe, queryable, and useful. That includes building the knowledge graph, the retrieval layer, the pipelines that keep the data current, and the access controls that ensure users only see what they are allowed to see. The role is not simply about document search with a graph label attached; it is intended to address graph-shaped problems such as fraud networks and payment chains from the outset.

The role may be placed either in the shared platform group or embedded within one product line. In the platform model, you would build the common knowledge layer used by both product lines. In the embedded model, you would own the system of record and domain data model for a specific line. The final shape would be agreed with you based on where your strengths are strongest.

What you will be responsible for

• Building and operating the production knowledge graph, including the ontology, schema, entity matching logic, and the pipelines that create and refresh the graph from client estates.
• Making graph retrieval a core capability for the platform, including traversal patterns, graph-augmented retrieval for agents, and query performance that holds up under real usage.
• Owning the retrieval stack end to end, from chunking and embedding choices through hybrid search and reranking, and defining the measurements that show whether a change genuinely improved results.
• Creating data pipelines that ingest information from client systems, map it into the platform, validate data quality, monitor freshness, and continue running unattended with alerts that are actually useful.
• Maintaining an immutable audit trail showing what the system read, what it decided, and what actions it took, along with the reporting views built on top of that trail.
• Implementing access controls that are aware of the caller’s identity, enforce document-level permissions, and make tenant isolation demonstrable rather than assumed.
• Designing the data layer that supports agent memory, including how episodic and precedent memory is written, retrieved, expires, and is governed.
• Defining data contracts between the pipelines owned by each product line and the shared retrieval plane so that the same entity is not modelled in conflicting ways across the organisation.
• Working in a senior engineering model where engineers produce specifications, harnesses, evaluation suites, and guardrails, while AI agents carry out most of the implementation work.
• Collaborating with quality and evaluation engineers, domain experts, and other senior team members to deliver regulated software through the team’s approval gates for specification, merge, and release.

What the company is looking for

• Strong hands-on production data engineering experience using Python, SQL, and a pipeline orchestration tool to run systems that other people depend on.
• Deep graph engineering knowledge, including ontology and schema design, experience with a production graph database such as Neo4j or something comparable, and confidence using a graph query language.
• Real-world entity resolution experience, including matching people, organisations, policies, accounts, and claims across source systems that do not agree with each other.
• Experience building retrieval systems for AI workloads, including embeddings, vector databases, hybrid search, reranking, and rigorous evaluation.
• Practical discipline around regulated data, including handling PII and PHI, lineage, retention, minimisation, and the fact that masking test data is essential.
• Comfort using AI coding assistants every day as part of your normal development process.
• Experience with systems of a similar shape on equivalent technologies is acceptable; the company does not require an exact match to the named tools in its reference stack.
• Ability to work in a hybrid pattern rather than a fully remote arrangement, with time split between offices, client sites, and home.
• Ability to pass pre-employment screening, including identity checks, nationality or immigration-status verification, three years of employment history, and an unspent criminal-record check.

Experience and knowledge that would help you stand out

• Work on graph problems in financial crime, fraud networks, AML investigations, or payment-chain analysis.
• Familiarity with a domain data model the company integrates with, such as ISO 20022, Guidewire, or a core banking or policy administration schema.
• Evidence that you have improved retrieval quality and can explain the quality curves you moved and the reasons behind the changes.
• Broader analytical data-platform experience, for example with ClickHouse or a similar system, alongside transactional and graph stores.
• A thoughtful approach to context engineering, treating the model context window as a managed engineering resource rather than an informal prompt-writing exercise.

Reference technology environment

The supported platform uses self-hosted LangSmith and LangGraph Platform for agent runtime and evaluation, a swappable gateway layer for model providers, PostgreSQL with pgvector, ClickHouse, and S3-compatible object storage for the data platform, Neo4j Enterprise for the semantic knowledge graph, an agent memory layer that serves episodic and precedent memory through MCP, MCP-native connectors, OpenTelemetry and Grafana for observability, and CNCF-aligned Kubernetes with Helm and Argo CD for deployment. The environment is designed to run on any hyperscaler or on-premises infrastructure.

An exact one-to-one match with that stack is not required. Comparable experience with different orchestration tools, graph engines, evaluation platforms, or serving layers is fully acceptable if you have built and operated systems with the same overall shape.

How the team works

Engineers are expected to write the specifications, harnesses, evaluation suites, and guardrails, while AI agents perform most of the implementation loops. Engineering judgement is applied through review rather than through manual typing. The team uses three human approval points for everything that ships: approval of the specification, approval of the merge, and approval of the release. Code paths that affect money movement, authentication, cryptography, or secrets remain under human ownership at all times.

The organisation is intentionally small and senior. There are no scrum masters and no separate delivery-management layer; quality is owned inside the product team by the engineers who build the software and the quality and evaluation engineers who work alongside them. Domain specialists, including claims practitioners and payment scheme experts, are embedded as full-time members of the teams.

What success looks like in the first year

• The graph build and refresh process runs unattended against a live client estate, with freshness actively measured and alerted on rather than simply assumed.
• Retrieval quality for at least one product line improves on a metric accepted by the evaluation engineers, and you can explain what drove the improvement.
• Permission-aware retrieval continues to hold up under adversarial testing, and multi-tenant isolation has been demonstrated to a client security team.
• A second product line reuses the data model and contracts you established instead of creating its own incompatible version.

Inclusion and recruitment approach

Capgemini states that it is a Disability Confident Employer at Level 2. As part of that commitment, candidates who declare a disability and meet the minimum essential criteria will be offered an interview, provided they opt in during the application process.

Rewards, wellbeing, and development

The company says it offers a total reward package that goes beyond pay and includes both core and flexible benefits. Employees also have access to the Applaud peer recognition portal. Capgemini says it was recognised as a Great Place to Work for Wellbeing in 2024 and emphasises the importance of employee wellbeing to the organisation’s ambitions.

Wellbeing support includes trained Mental Health Champions across the business and access to wellbeing apps such as Thrive and Peppy. The company also highlights a strong learning culture, with opportunities to explore, innovate, and progress, plus extensive development options ranging from thinktanks and hackathons to a large catalogue of courses and external certifications from providers such as AWS, Microsoft, Harvard ManageMentor, and cybersecurity organisations.

About the wider organisation

Capgemini describes itself as an AI-powered global business and technology transformation partner that delivers business value through AI, technology, and people. The group says it has nearly 60 years of heritage, operates in more than 50 countries, and has around 420,000 team members. Its work spans strategy, technology, design, engineering, and business operations, supported by a broad partner ecosystem. The company says it helps clients grow, innovate, and reimagine what is possible through technology.

The role is a permanent position based in London, and the posting identifies the experience level as experienced professional. The professional community for the role is software engineering.

To apply, candidates should use the application process provided on the job page, including the option to apply directly or start an application through LinkedIn.

---

Degree Requirement: Not Specified

Visa Sponsorship May be