AI Engineer – Generative AI & Agents

Role purpose

Build generative AI applications and agent-based workflows that support employees and improve brokerage operations. Develop reliable retrieval, reasoning and tool-use capabilities around approved company knowledge and data. Turn use cases such as support assistance, policy search and operational investigation into tested, observable and maintainable software.

Key responsibilities

AI application development

• Develop Python services and APIs for internal assistants, client-support copilots, document processing, case summarisation and operational analysis.
• Build retrieval-augmented generation pipelines using approved policies, procedures, product specifications and knowledge content, with source attribution and document version awareness.
• Implement agent workflows with structured outputs, tool schemas, bounded execution, state management, approval checkpoints and clear termination conditions.
• Develop brokerage-specific responses that respect entity, product, language and client context; prevent unsupported claims about trading conditions or account status.

Quality and model optimisation

• Compare model and retrieval approaches using representative datasets; select configurations based on grounded accuracy, task completion, latency and operating cost.
• Improve chunking, embeddings, metadata filters, reranking and context construction while preserving source permissions and preventing cross-client information leakage.
• Maintain versioned prompts, model configurations and evaluation datasets; use fine-tuning only where a validated business case and suitable approved data exist.
• Work with QA to test hallucinations, prompt injection, sensitive information disclosure, ambiguous instructions and failure to request human assistance.

Production delivery and support

• Integrate approved enterprise tools through controlled interfaces, using authenticated user context and explicit authorisation for actions that modify records or trigger workflows.
• Instrument application traces, retrieval results, model usage and tool outcomes with appropriate redaction, enabling diagnosis without exposing unnecessary client information.
• Implement graceful failure, timeouts, retries and human handoff when information is missing, confidence is insufficient or a requested action exceeds permissions.
• Maintain code, tests and operational documentation; investigate production failures and use reviewed feedback to improve application performance over time.

Essential experience and qualifications

• 4+ years in software, data or ML engineering, including at least 2 years building applied NLP, generative AI or LLM-based systems.
• Evidence of shipping and supporting production AI applications, including retrieval or agent workflows; prompt-only experimentation is insufficient.
• Strong Python, API development, automated testing, Git and software design skills.
• Hands-on experience with LLM APIs or open-weight models, embeddings, vector or hybrid search, structured outputs and tool calling.
• Understanding of evaluation design, retrieval quality, prompt injection risks, context limitations and the variability of model outputs.
• Ability to work with SQL, enterprise documents and permission-controlled data and to explain application limitations to business stakeholders.
• Bachelor’s degree in Computer Science, AI, Engineering or a related field, or equivalent practical experience.

Preferred experience

• Financial services experience, particularly brokerage operations, trading support or client service.
• Experience with an agent orchestration framework, multilingual retrieval and document extraction or OCR.
• Experience with model adaptation, inference optimisation and observability for LLM applications.