岗位描述
Design and build agentic AI solutions (multi-step reasoning, tool-use, orchestration, memory, and guardrails) that reliably execute business workflows. Develop production-grade Python services and pipelines that integrate LLMs with internal systems, APIs, data assets, and controls. Implement tool/function calling, retrieval (RAG), evaluation harnesses, and prompt/agent optimisation to improve accuracy and reduce risk. Partner with domain stakeholders to identify high-value opportunities, define success metrics, and deliver solutions that demonstrably improve outcomes. Apply financial-services discipline: security-by-design, data minimisation, auditability, model risk considerations, and operational resilience. Use coding assistants effectively (e.g., GitHub Copilot, OpenCode or similar) to accelerate delivery while maintaining high engineering standards. Contribute reusable components (agent templates, tool registries, eval suites, reference architectures) to scale delivery across teams. Support deployment and operations: CI/CD, observability, model/agent telemetry, incident response, and iterative enhancement. Proven hands-on experience building and shipping agentic AI solutions (not just experimentation), including at least one successful delivery of an AI solution into a real user workflow. Strong Python engineering skills (API development, data processing, testing, packaging) and comfort building maintainable, reviewed code. Practical experience with modern LLM frameworks/patterns (e.g., tool calling, RAG, planning/execution loops, routing, evals), and knowing when not to use agents. Solid understanding of financial-services context and constraints (risk/compliance mindset, privacy, security, auditability, regulatory awareness). Experience integrating AI into enterprise systems (REST APIs, data stores, event-driven patterns, authentication/authorisation). Ability to work end-to-end: requirements → design → build → test → deploy → measure impact. Strong communication skills—able to translate business problems into technical approaches and explain trade-offs clearly.