Expert: Agentic & MCP Architect

MoMo from MTN; Johannesburg; On-site; ZAR 1,200,000 - 2,100,000; Full time

Job summary

MTN Group Fintech is building an AI capability from the ground up to power financial services across markets. The Agentic & MCP Architect will own agent architecture, tool integration, and reusable execution patterns, shaping how autonomous agents interact with enterprise systems.

This is a hands-on architecture role within a build team, designing the agent framework and setting technical direction for engineers who build production agents. Closing date: 01 October 2026.

Qualifications

• Bachelor’s degree in Computer Science, Engineering, Data Science, Mathematics or a related technical field.
• Minimum of 8 to 10 years in software or machine learning engineering.
• At least 3 years in a technical design authority or lead engineering role.
• Hands-on design and build of agentic/workflow automation systems in production.
• Experience with large language models, agent frameworks, tool and function calling, context protocols, memory design.

Responsibilities

• Own agent architecture, including planning, memory, orchestration and multi-agent design.
• Oversee tool and context integration layer, including Model Context Protocol servers and connectors.
• Develop reusable execution patterns, agent templates and the agent runtime.
• Design safe action mechanisms with permissioning, human-in-the-loop, escalation and rollback.
• Define agent evaluation design and ensure behavioural quality in production.
• Align agent patterns with enterprise AI architecture, security, data and risk frameworks.
• Escalate non-standard patterns for governance approval.

Skills

Agent architecture

Memory design

Orchestration

Multi-agent design

Agent framework

Agent runtime

Safety and permissions

Human-in-the-loop

Governance and risk awareness

Production-grade design

Education

Bachelor's degree in Computer Science, Engineering, Data Science, Mathematics or related technical field

Job description

MTN Group Fintech is building a world-class AI capability from the ground up, creating the platforms, agents, and intelligent systems that will power the next generation of financial services across multiple markets. This is a rare opportunity to join a newly established AI squad at the start of its journey and help define how AI transforms customer experiences, operations, growth, risk management, and productivity across one of Africa's largest fintech ecosystems.

The Agentic & MCP Architect is the design authority for how Group Fintech builds autonomous and semi-autonomous agents. The role owns agent architecture, the tool and context integration layer, and the reusable execution patterns through which agents act on enterprise systems safely and predictably.

This is a hands-on architecture role within a build team. The Agentic & MCP Architect designs and builds the agent framework and tool servers, writes the patterns, reviews agent implementations, and sets the technical direction for the engineers who build production agents. The role does not direct delivery by consultancies or vendors.

As part of your portfolio as Expert: Agentic & MCP Architect, you will be accountable for.

• Agent architecture, including planning, memory, orchestration and multi-agent design
• The tool and context integration layer, including Model Context Protocol servers and connectors
• Reusable execution patterns, agent templates and the agent runtime
• Safe action design, including permissioning, human-in-the-loop, escalation and rollback
• Agent evaluation design and behavioural quality in production

Agent design decisions taken in this role determine what the enterprise can safely automate and set the pattern that every agent engineer builds to. The role works within the enterprise AI architecture and the security, data and risk frameworks set by Technology & Platforms and Risk, and escalates new or non-standard patterns for governance approval.

The incumbent must have the following:

• Bachelor’s degree in Computer Science, Engineering, Data Science, Mathematics or a related technical field.
• Minimum of 8 to 10 years in software or machine learning engineering, including at least 3 years in a technical design authority or lead engineering role.
• Demonstrated hands-on design and build of agentic or workflow automation systems running in production.
• Practical depth in large language models, agent frameworks, tool and function calling, context protocols, planning and memory design.
• Experience integrating automated systems with enterprise applications and core transactional platforms.
• Experience designing permissioning, guardrails and human oversight for systems that take action.
• Experience defining engineering standards and patterns consumed by other engineers.
• Exposure to regulated environments and to security and risk review of automated decisioning or action.

What will give an edge to your application:

• Recognised cloud, security or AI engineering certification.
• Financial services, fintech or another high-volume transactional environment.

Closing date: 01 October 2026. Late applications will not be accepted.

Should you not hear from us within 2 weeks of closing date, please consider your application unsuccessful