Role Overview
Agency AI, a UK-based AI consultancy, is seeking a skilled Senior AI Engineer to join our team on a contract basis. You will play a pivotal role in developing an advanced AI agent system to replace a legacy OCR-based order-capture system. This system will enhance our client's capabilities by reading customer order emails in various formats and applying complex business rules to process orders accurately.
ResponsibilitiesA
- Develop an agentic system using Python and Claude agentic frameworks, focusing on prompt caching, tool definitions, memory, evaluation frameworks, structured outputs, and a deterministic pre-filtering layer.
- Integrate with Azure Content Understanding or similar services for extracting information from multi-format documents.
- Construct a rules engine to encode over 20 distinct business tools, including customer matching, address resolution, item-code translation, and pricing validation.
- Implement REST API integration with third-party tools.
- Create a lightweight web-based review interface in React for human-in-the-loop verification.
- Develop an evaluation harness to measure agent accuracy against historical orders.
- Deploy the system in the client's Azure tenant with logging, cost-monitoring, and spend caps.
- Focus on cost-engineering to optimize system operations.
Required Skills
- Proven experience as a senior professional software engineer with a track record of deploying production systems utilizing large language model APIs such as Anthropic or OpenAI.
- Proficiency in Python.
- Demonstrated expertise in agentic patterns, including tool calling, prompt caching, structured outputs, retrieval, and evaluations.
- Experience in integrating with enterprise REST APIs, managing OAuth2, and handling validation flows and production-data writes.
- Excellent written English skills for clear communication with non-technical stakeholders.
Nice to Have
- Experience with Microsoft Dynamics 365 Business Central, NAV, or other enterprise ERP integrations.
- Familiarity with Azure services, especially Content Understanding or Document Intelligence.
- Experience in building human-in-the-loop review interfaces and evaluation harnesses.
- Knowledge of prompt caching strategies and cost optimization for LLM-based systems.