Description
The role operates within established architectural direction and is accountable for delivering reliable, maintainable, and production ready solutions that meet business, security, and compliance requirements.
The Role
- Design, build, and test multi‑step agent workflows using established patterns such as ReAct, planner‑executor, and tool‑chaining
- Integrate LLMs (i.e. Claude and OpenAI) with enterprise APIs and internal systems, including robust handling of retries, edge cases, and degraded states
- Implement tool calling, function orchestration, and compensating actions to ensure workflows remain stable under failure conditions
- Design and implement human‑in‑the‑loop controls, including approvals, escalations, and exception handling where required by business logic or risk considerations
- Collaborate with product managers, domain experts, and technical stakeholders to refine requirements and translate them into technical designs
- Develop, maintain, and version prompt logic and supporting documentation in line with agreed standards
- Build and manage agent memory and retrieval mechanisms using vector databases and retrieval‑augmented generation (RAG), ensuring context remains relevant and high‑signal
- Deploy AI services to production environments and actively monitor logs, traces, and metrics to detect and resolve issues
- Contribute to test coverage, operational runbooks, and incident response practices in collaboration with QA and operations teams
- Support continuous improvement of system reliability, performance, and maintainability
Qualifications
- Minimum 2 years’ experience developing backend or service-based systems using C# and/or Python
- At least 1 year of hands‑on experience working with large language models, prompt engineering, or AI‑enabled systems
- Strong analytical and debugging skills with the ability to reason across APIs, data flows, and AI system behavior
- Ability to clearly communicate technical decisions and trade‑offs to technical and non‑technical stakeholders
- Demonstrated ability to learn and adapt in a rapidly evolving technology environment
- Programming: C#, Python (production services, async workflows, automated testing)
- Cloud Platform: Azure (compute, storage, identity, deployment pipelines)
- LLM Integration: Claude and/or OpenAI APIs (prompting, tool calling, rate limits, error handling)
- Agent Orchestration: Azure AI Foundry, Microsoft Agent Framework; optional alternatives include LangGraph and LangChain.
- Vector Databases: Pinecone or equivalent (indexing, retrieval, relevance tuning)
- APIs & Integration: RESTful APIs, enterprise authentication, service‑to‑service integration
- Observability: Application logging, metrics, and tracing in production environments
WTW is an Equal Opportunity Employer