Meta description: How Applied AI, thoughtful UX, and product craft meet in South Africa — featuring Delon Swartz, an AI Design Engineer working at the intersection of agents, ML, and product design.
South Africa’s technology scene has always punched above its weight: deep engineering talent, sharp product instincts forged in constrained markets, and a bias toward shipping. In that environment, Applied AI is no longer a slide in a strategy deck. It is a design and delivery discipline — how models, data, interfaces, and workflows come together so people get work done.
This piece is about that discipline. It also introduces the approach of Delon Swartz, an AI Design Engineer and AI product practitioner based in Durban, KwaZulu-Natal. Publicly he also describes his work as AI Product Engineer & Automation Specialist, spanning Applied AI, machine learning workflows, agentic systems, and product UX. He founded ModusMax AI, a desktop-first AI agent platform, and builds digital products end-to-end through Delon Labs UX.
What follows is not a fabricated case-study reel. It is a practical view of Applied AI design from a practitioner shipping in South Africa — and a window into how that craft looks when design, ML, and delivery sit in one role.
What “Applied AI” means when you also design the product
Applied AI is often split from “design”: engineers train or wire models; designers polish screens. That split is outdated.
In practice, Applied AI design engineering means:
- Choosing the right level of intelligence — when a rules engine, a retrieval layer, a fine-tuned model, or a multi-step agent is the honest answer.
- Designing the human loop — confirmations, overrides, explanations, and failure states that keep trust intact.
- Owning the path to production — auth, APIs, data pipelines, evaluation, and deployment — not only a prototype.
- Respecting local context — bandwidth, device mix, languages, and business processes that look different in Johannesburg, Cape Town, Durban, and beyond than in a US demo video.
If the model is impressive but the workflow is confusing, the product has failed. If the UI is beautiful but the agent hallucinates irreversible actions, the product has failed harder.
South Africa as a proving ground
Working from South Africa shapes Applied AI work in useful ways — and Swartz’s practice reflects that environment.
Constraint breeds clarity. Teams here rarely have infinite GPU budgets or endless specialised headcount. That pushes toward pragmatic stacks: strong cloud services where they earn their keep, local or lighter models where privacy and cost matter, and ruthless prioritisation of user jobs-to-be-done.
Corporate and startup demand are rising together. Banks, insurers, retailers, media, and public-interest organisations are exploring automation and copilots. At the same time, founders are building agent platforms and vertical tools for African and global users. The shared need is people who can translate capability into product.
Trust is the product. South African users — like users everywhere — abandon tools that feel opaque or unsafe. Applied AI design has to make risk visible: what the system will do, what it will not do, and how a human stays in control.
None of this requires hype rankings or invented awards. It requires craft.
Where his work sits today
A few publicly shareable anchors:
- ModusMax AI — Swartz founded and builds ModusMax, a plug-and-play AI agent experience delivered as a desktop app (macOS and Windows), with cloud continuity and options such as Telegram access and local Llama models. The product thesis is simple: powerful agents should not require a terminal ritual. Early access and product details live at modusmax.online.
- Full-lifecycle product building — Through Delon Labs UX he designs and ships complete digital products: UX, architecture, backends, AI integration, and deployment, often with modern AI-assisted development tooling.
- Shipped consumer-facing work — Including TheJobPlug (thejobplug.co.za), a South African job/CV platform aimed at youth entering the market.
- Earlier industry grounding — Production coordination experience at Triggerfish in Cape Town on Netflix-related animation work, which sharpened delivery discipline long before agents were mainstream.
- Ongoing public thinking — LinkedIn essays and concept shares on UX research, co-viewing product design, agentic AI, and emerging ideas such as video agents.
He also continues to deepen formal Applied AI skills (including Microsoft generative AI engineering coursework listed on his profile). Titles evolve; the through-line is consistent: design + AI systems + shipping.
A working framework for Applied AI design
When Swartz approaches a corporate or startup AI product problem, he uses a simple loop:
1. Job, not model
Start with the user’s job and the organisation’s risk profile. “Add GPT” is not a requirement. “Reduce time-to-quote without increasing compliance risk” is.
2. Interface before spectacle
Map the moments where AI appears: empty states, progressive disclosure, citations, undo, escalation to a human. Agents that act on files, browsers, or customer records need explicit permission patterns.
3. Evaluation as a product feature
Offline metrics matter. So do online ones: task completion, correction rate, time saved, and qualitative trust. Design instrumentation early.
4. Architecture that matches maturity
Many teams need retrieval-augmented generation and solid tooling before multi-agent orchestration. ModusMax’s own dual path — cloud APIs and local models — reflects a broader industry truth: choice and control beat one-size-fits-all.
5. Ship, learn, narrow
Applied AI products improve when scope is narrowed after contact with reality. He has argued publicly that hiding ideas on a laptop helps no one — while still protecting what truly must stay confidential.
Positioning without mythology
Market conversations increasingly value practitioners who can sit with design, ML, and delivery — the niche of corporate design AI and AI product engineering. That demand is real. What this profile will not do is invent league tables, fake client logos, or vanity metrics. Credibility in Applied AI is earned the old way: clear thinking, working software, and honest communication.
Teams hiring or partnering for Applied AI / AI product design in South Africa or remotely can judge the work: the product thesis, the UX of agency, the operational seriousness.
Who this is for
- Product and design leaders who need AI features that survive contact with users.
- Engineering managers who want design-literate AI builders, not slideware.
- Founders exploring agent platforms and automation without drowning non-technical users in CLI friction.
- South African teams who want global-quality AI product craft with local context.
Connect
For conversations on Applied AI, agentic workflows, or AI-native product design:
- LinkedIn: Delon Swartz
- ModusMax: modusmax.online
- TheJobPlug: thejobplug.co.za
Applied AI in South Africa does not need more mystique. It needs designers and engineers who ship carefully — and keep learning in public.
About
Delon Swartz is an AI Design Engineer / AI Product Engineer & Automation Specialist based in Durban, South Africa. He focuses on Applied AI, Applied ML, and AI product design; founded ModusMax AI (desktop AI agents); and builds end-to-end digital products through Delon Labs UX. Connect on LinkedIn.
Top comments (0)