Senior Agentic AI Engineer

Deviare; South Africa; On-site; ZAR 900,000 - 1,500,000; Full time

Job summary

Deviare is seeking a Senior Agentic AI Engineer in Johannesburg to design, build and operate AI agents that perform real business tasks. You will work on single- and multi-agent systems with tool-calling and enterprise integration, ensuring identity, governance and observability from day one.

The role requires hands-on development, memory/orchestration patterns and experience with MCP/A2A protocols, along with strong security and risk controls in a regulated financial services context.

Qualifications

• Strong Python programming with API integration experience.
• Experience delivering agentic solutions in production environments.
• Familiarity with security, governance and observability in regulated environments.

Skills

Python programming

API integration

LLM agents

Enterprise integration

Education

Bachelor’s or Master’s degree in Computer Science or Engineering

Tools

Google ADK

LangGraph

Semantic Kernel

CrewAI

AutoGen

Docker

Kubernetes

MLflow

Datadog

OpenTelemetry

Job description

We are seeking a Senior Agentic AI Engineer to join a leading Financial Services and Insurance organisation in Johannesburg. This permanent role starts immediately and reports to the Head of AI / Engineering Lead. The role involves designing, building and operating AI agents that carry out real business tasks. These are single- and multi-agent systems that plan, reason, use tools and act across enterprise systems. Security, agent identity, governance and observability are built in from the start.

Context & Impact

In a regulated financial services and insurance environment, AI agents that take actions carry much higher stakes than models that only generate content. Every action an agent takes in an enterprise system has to be authorised, traceable and reversible where needed. Every agent needs a defined identity with least-privilege access.

This role builds agents that integrate with enterprise systems and APIs through tool-calling and protocols such as MCP and A2A. It covers memory, orchestration and human-in-the-loop approvals. The role also puts the controls around those agents in place: protection against prompt injection and unintended actions; evaluations of task success, tool use and safety; GRC controls aligned with POPIA, FSCA/PA regulations and internal model-risk policies; and end-to-end observability. The Senior Agentic AI Engineer works with the business to identify high‑value processes for automation. They also help define the organisation’s agent engineering and governance standards.

Why This Role Is Exciting

• Agents that act: build single- and multi-agent systems that plan, use tools and complete end‑to‑end business tasks in production.
• Enterprise integration: connect agents to core systems and APIs using tool-calling and emerging protocols such as MCP and A2A.
• Agent identity and security: design least‑privilege access, service identities and guardrails against prompt injection and unintended actions.
• Agent evaluation: measure task success, evaluate tool‑use trajectories and trace agent behaviour, alongside red‑teaming and regression testing.
• Setting the standard: help define how agents are versioned, registered, approved and governed across the organisation.

What We’re Looking For

The ideal candidate has at least 7 years of software engineering experience, including 2+ years building LLM‑based agents or automation. They have delivered agentic solutions into production. Financial services experience is an advantage.

Core strengths include:

• Expert Python, with strong API and integration experience
• Agent frameworks such as Google ADK, LangGraph, Semantic Kernel, CrewAI or AutoGen
• Agent patterns: tool calling, planning, memory, multi‑agent orchestration and MCP/A2A protocols
• Identity and security for agents: OAuth and service identities, RBAC and secrets management
• Azure, AWS or GCP, with Docker and Kubernetes
• AI evaluation, observability (MLflow, Datadog, OpenTelemetry) and governance, including agent lifecycle controls and audit readiness

The right person treats autonomy and control as equally important design goals. They think carefully about what an agent should and shouldn’t be allowed to do. They give honest advice on feasibility, risk and controls, and they are comfortable in a fast‑moving field where standards are still forming. They also mentor engineers and help shape good practice. A Bachelor’s or Master’s degree in Computer Science, Engineering or a related field is required. Google Cloud, Azure AI Engineer or AWS AI/ML certification is an advantage.