External: Agentic AI Developer

About the company
Our client is a well-established South African fintech in the payments space, building AI agents to automate work across the business.

About the role
Join a team building AI agents that take on real work across operations, support, fraud, risk and engineering. You'll build, test and ship agent components alongside senior developers, and learn how to put AI safely into production in a regulated payments environment. You'll be expected to ship at least one agent feature to production, with evaluation tests in place, within your first six months.

Key responsibilities

• Work with product, operations and risk teams to understand the processes an agent will support
• Build and maintain agent workflows, including tool calling, RAG, prompt design and multi-step task orchestration
• Connect agents to internal APIs and data sources through secure, well-scoped integrations
• Write evaluation tests that check agent output for accuracy, reliability and safety before release
• Monitor agents in production, triage failures and feed findings back into improvements
• Document what each agent can and cannot do, and where a human signs off
• Follow change, release and security processes for every AI feature, and handle personal and financial data in line with POPIA
• Help business users adopt new agent tools through guides and demos
• Keep up with agent frameworks, MCP and model releases, and propose new automation use cases backed by the problem they solve

Ideal candidate

• A degree in Computer Science, Engineering or a related field, or equivalent practical experience
• 2+ years of software engineering experience, including internships or strong project work
• Solid Python, and comfortable with REST APIs and JSON
• Hands-on experience building with at least one LLM API, shown through a project, GitHub repo or production work
• A basic understanding of RAG, embeddings and prompt engineering
• Git and working knowledge of Docker
• A careful approach to personal and financial data

Advantageous

• Exposure to an agent framework (such as LangGraph, CrewAI or an agent SDK) or to MCP
• AWS, C# or .NET, and SQL (SQL Server or PostgreSQL)
• An interest in payments, fraud or financial services