The Role
You will join the AI Team in Front Office Technology to design, build and maintain the firm’s knowledge layer. Its main users are investment teams and the AI tools that support them. It also covers the technology estate, so engineering and operations tools share the same picture. You will work closely with our AI Engineers, the wider engineering community and DevOps so the layer is maintained as part of everyday work, largely through agents.
This is a hands-on engineering role. Data engineering moves and stores data; this role defines what things are, how they relate and where knowledge came from, so people and AI agents can use it.
Role Purpose
What the firm knows, and how that knowledge connects, is a large part of what sets it apart. Making it usable by AI multiplies its value, with people and tools then able to build on it.
Significant information already exists across the firm: reference data, research, documents and system records. Documentation can sometimes be outdated or lacking, and the connections between the things we use, such as issuers, bonds, strategies, systems and services, are often not defined or maintained.
Without them, AI tools can read each source on its own but cannot reliably move across them. AI will determine links for itself, at times slightly differently, with answers therefore becoming harder to trust or audit.
A knowledge layer addresses these problems. It gives each thing a standard description and maps how things relate, so AI can follow those links to answer questions and carry out tasks across systems. It is not a one-off project: it needs continuous maintenance, much of it carried out by agents with people confirming changes. For example:
- A PM asks how a sovereign downgrade could affect their book. The answer spans the country’s bonds, rates curves and currency, state-owned companies whose debt trades with the sovereign, and the latest research on all of them. No single system holds those links, sources describe them differently, and some exist only in documents or people’s experience. The knowledge layer records them so an agent can follow them, within the user’s entitlements.
- An agent investigating a failing service can find what depends on it, who owns it and its runbook.
- Any answer can be traced back through those links to its sources.
How We Work
- Thin and useful. We model what is needed, when it is needed.
- Agentic tool calling. Agents use well-defined tools with access control built in.
- Deterministic where it counts. Where accuracy matters, we use deterministic methods, not model recall.
- Maintained by agents, confirmed by people. Agents do the routine upkeep; people confirm what changes.
- Teams stay in control. Teams own their knowledge and decide what is shared beyond them.
Key Responsibilities
Design
- Design the knowledge model across front office and technology domains: instruments, issuers, research, themes, systems, services and owners, and how they relate.
- Manage the model as code, with version control, automated validation and clear change management, so others can contribute to and support it.
Build
- Link knowledge reliably to existing reference data, research, documents and system records, starting with the most reliable sources.
- Design taxonomies and tagging for research, commentary, internal notes and runbooks.
- Build a small set of well-defined tools that AI agents use to find entities, retrieve knowledge and follow relationships, with access control enforced inside them.
Support teams
- Help teams, including front office desks, define, capture, structure and share their knowledge, with simple tools standards, linked notes that people and agents can both read and update, and agent-assisted tools, demonstrating value early.
- Make capture part of everyday work, as notes, research and runbooks are written, not a separate task through the use of tooling built natively into emerging AI oriented workflows and with the support of our AI Engineers.
Maintain
- Build agents that keep the layer current: detect changes in source systems and code, propose updates and new links, and flag stale or conflicting knowledge for review.
- Build updates into normal engineering work, for example when a service, pipeline or dataset changes.
- Record provenance and full history for audit, and manage change without breaking dependent tools.
- Measure coverage, accuracy and effect on AI answers, and test graph technologies on real use cases.
Essential
Key Experience and Qualifications
- 5+ years in data engineering, data modelling or knowledge engineering, with strong Python and SQL, and daily use of AI coding tools.
- Proven knowledge modelling or ontology design: entity and relationship models that other teams built on in production.
- Hands-on experience with knowledge graphs, using at least one graph technology (property graph or RDF).
- Experience structuring unstructured content: taxonomies, tagging and document metadata.
- Experience designing typed APIs or tool integrations that AI agents and other systems depend on.
- Hands-on experience building LLM agents and grounding them in structured knowledge through tool use, retrieval and structured outputs, and measuring their quality.
- Schema governance (versioning, automated validation, controlled change) and point-in-time modelling.
- Sound judgement on what to model now and what can wait, and the ability to explain trade-offs clearly.
Desirable
- Specific graph platforms (e.g. Neo4j/Cypher, SPARQL stores) and graph-based retrieval.
- Building MCP servers; Snowflake and semantic layers; industry ontologies such as FIBO.
- Technology operations data: service catalogues, CMDBs, observability or runbooks.
- Financial reference data (e.g. FIGI, ISIN, LEI) and Rates or EM knowledge, ideally from a hedge fund or bank.
BlueCrest is committed to providing an inclusive environment for its workforce. As an employer, we provide equal opportunities to all people regardless of their gender, marital or civil partnership status, race, religion or ethnicity, disability, age, sexual orientation or nationality.
#J-18808-Ljbffr…