Machine Learning Engineer

Location: Johannesburg, South Africa

Department: Technical Team

You're not joining a QA team. You're joining a mission to make broken software a thing of the past.

At Duku AI, we're building the next generation of autonomous testing agents — systems smart enough to understand what your product is supposed to do, find where it fails, and fix themselves when the codebase changes. No flaky test scripts. No brittle selectors. No 3am pages because a deployment broke something no human thought to check.

As a core member of our team, you'll work at the intersection of AI systems, developer tooling, and product engineering. You'll own meaningful problems end-to-end — from shaping how our agents reason about software behaviour, to ensuring our platform scales across the world's most demanding engineering teams.

This is a high-autonomy role. You'll collaborate directly with founders who have built and scaled infrastructure at Meta, Uber, and Deliveroo. You'll move fast, think deeply, and ship things that matter.

What you'll do:

• Design and build systems that make autonomous software testing a reality at scale

• Work closely with engineering and product to define how AI agents interact with real-world codebases

• Own your work from first principles through to production — with full context and full accountability

• Shape the culture, processes, and technical standards of a team that's serious about craft

What we're looking for:

• Deep technical ability and an instinct for elegant solutions to hard problems

• Someone who raises the bar around them — for quality, for thinking, for execution

• A builder who thrives with ambiguity and ships with conviction

• Curiosity about AI, autonomous systems, and the future of developer tooling

If you've ever been frustrated watching slow, manual QA block great engineering — this is your chance to fix it.

About the Company

Change Software Forever

Most ML roles optimise models in isolation. This one builds the intelligence layer that determines whether AI-generated software can actually be trusted in production.

Why this matters now

AI is rewriting how software gets built. But shipping still breaks on testing. Engineering leaders are desperate to release faster without dropping quality, and they can't get there with manual QA or brittle test automation.

Here's the bigger shift: as AI writes more of the code, the bottleneck moves from generating software to trusting it. Nobody owns that trust layer yet. Duku is building it, autonomous agents that simulate real users, catch critical failures before production, and self-heal as the product evolves.

Backed by experienced operators and investors, we're building what we believe will become a core layer of the modern software stack. We're looking for an exceptional Machine Learning Engineer to help build it. Not someone to tune existing models. Someone to design new ones.