Healthcare Jobs in Cape Town, South Africa (Now Hiring) — 89 open — Page 4

AI Builder (6-Month Contract)

Best10 | Remote | Reports to CEO, working alongside our Product Manager

Best10 delivers highly personalised diet and lifestyle coaching at scale through WhatsApp and our own technology platform. We partner with leading health insurers and employers to help people build healthier habits, lose weight sustainably and reduce their long-term risk of chronic disease.

Our biggest challenge isn't creating great coaching, it's making great coaching scalable.

Today, each of our coaches supports around 200 members. Our ambition is to dramatically increase that number without compromising the personal, human experience that makes our programme so effective. In fact, our goal is even more ambitious than that: every improvement we make should allow a coach to support more members while simultaneously improving the quality of care every member receives.

We believe technology and AI should never replace the relationship between coach and member. Instead, they should remove administrative work, surface the right information at the right time, identify members who need support before they ask for it, and make every interaction more personalised and more impactful.

The Role

We are looking for a builder, not an analyst. Someone who is excited about finding gaps in the business, validating them with data, and shipping working AI solutions in days, not sprints.

This is not a platform or strategy role. You will embed with our coaching team, understand what a day at 1:200 actually looks like, and build the AI workflows that let coaches spend their attention where it matters most. Everything you build must work for non-technical users and must survive after your contract ends.

You will work alongside our Product Manager, who owns the domain definitions (what "at risk" means, how coaches work) and who will run and extend these systems after your engagement. Your job is to build the engine. Theirs is to own it.

Your First 90 Days

We already know the first two problems. You will start here:

1. Risk ranking (months 1-3). Build the layer that answers "of my 200 clients, which 20 need me most today?" Disengagement signals, missed check-ins, sentiment shifts, adherence patterns. Most of this signal lives in unstructured WhatsApp conversations and platform data. Turning that into a reliable daily priority list is the core of the job.
    
2. Context briefings (months 3-5). When a client surfaces, the coach should have an instant briefing: goals, recent history, current risk, what has worked before. No scrolling through months of chat history.
    

Month 6 is handover: documentation, training, and making sure the systems run without you.

Key Responsibilities

• Build internal tools and coach-facing features on top of LLMs (Claude, OpenAI, open-source models)
• Turn unstructured conversational data (WhatsApp threads, coaching notes) into structured, reliable signal
• Design and maintain prompt chains, RAG pipelines, and agent workflows where they genuinely help
• Prototype rapidly, going from idea to working demo in days
• Evaluate model outputs, set up evals, and iterate on quality, with a focus on accuracy coaches can trust
• Wire AI into our existing systems (platform APIs, databases, WhatsApp, Slack)
• Own the full stack of what you ship: frontend, backend, model integration, deployment
• Sit with coaches regularly. Watch how they work. Build for their reality, not for a spec

What You Will Bring

• Demonstrable proof of using AI to build tools people actually use. We will ask you to show us, not tell us
• Proven experience shipping AI-powered products used by non-technical audiences
• Hands-on experience with LLM APIs (function calling, streaming, embeddings) and prompt engineering patterns
• Comfort working with messy, unstructured conversational data
• Strong product instinct: able to talk to a coach or founder, understand the real need, and build the right thing without a heavy spec
• Data-driven: willing to dig into the data, often building new reporting, to find root causes worth solving
• Fluency with agentic development tools (Claude Code, Cursor or similar) as part of how you work
• A builder's attitude toward handover: you take pride in systems that outlast you
• Fluent English

What Success Looks Like at Month 6

We will judge this contract on three things:

1. The risk ranking and context briefing workflows are live and used by coaches weekly, without prompting
2. Measurable improvement in time-to-intervention for clients who are disengaging or at risk
3. Our Product Manager can run and extend the systems independently

If those three are true, there is a strong likelihood this becomes a longer-term role shaping how AI runs through the whole business.

What This Role Is Not

• Not a dashboard or BI role. Reporting may be a byproduct, never the product
• Not an ML research role. No model training from scratch, no fine-tuning for its own sake
• Not a strategy role. We need working software in front of coaches, fast.